EMBEDDED SYSTEMS - OpenSystems Media

Daily Briefing: News SnippetsMilitaryEMBEDDED SYSTEMSVOLUME 6 NUMBER 3MAY 2010INCLUDING:Chris A. CiufoEmbedded Systems Conference portendsCOTS’ futureField IntelligenceGPUs close in on supercomputingMil Tech InsiderNAS expands security, storageLegacy Software MigrationMDA offers promise for legacy appsMIL-EMBEDDED.COMGIG: Our modernRed Ball ExpressMoving data,video, and ISRAlso:Battlefield security,from the ground up

MilitaryEMBEDDED SYSTEMSColumnsField Intelligence8 Clustered GPUs closing in onsupercomputer performanceBy Duncan YoungMil Tech Insider9 Networked Attached Storage expands datastorage and security on military platformsBy Steve EdwardsLegacy Software Migration12 Model Driven Architecture offers promisefor migrating legacy softwareBy D.K. McKean, Advanced Fusion TechnologiesCrosshairs Editorial46 Embedded Systems Conference portendsCOTS’ futureBy Chris A. CiufodepaRTMENTS14-15 Daily Briefing: News SnippetsBy Sharon Schnakenburg-Hess42-45 Editor’s Choice ProductsON THE COVER:During the World War II invasion of France, the Allies pressed southand eastward only by keeping the materiel flowing towards Germany. The6x6 “deuce” trucks did the lion’s share of hauling ammo, rations, and gasoline(often in the same load!). Dubbed the “Red Ball Express,” these vehicles formedthe backbone of Montgomery’s and Patton’s marches into Nazi territory. Today’smilitary relies on the Internet as the modern backbone. Net-centric COTSstandards form the battlefield’s Global Information Grid (GIG). Refer to articlesstarting on page 24.Published by:May 2010 Volume 6 Number 3ISSN: Print 1557-3222All registered brands and trademarks within Military Embedded Systems magazineare the property of their respective owners.© 2010 OpenSystems Media © 2010 Military Embedded SystemsSoftware: Nano-kernel commandsfast deployment16 POSIX RTOS attacks SWaP andtime-to-deployment issuesBy Kim Rowe, RoweBots Research Inc.Special: FPGAS and IP for video processing20 FPGAs balance architecture, IP, power inelectro-optical/infrared systemsBy Dr. Tibor Kozek, Imagize LLC; Juju Joyce, and Suhel Dhanani,Altera CorporationMilitary Materiel: IT and the GiG –Front and center24 Telecom equipment nets big benefits onthe battlefieldQ&A with Anthony Ambrose, VP and GM of RadiSys27 Designers are demanding end-to-end sensorprocessing, right out of the boxAn interview with Dipak Roy, Chairman of D-TA Systems30 Elements of a deployed, modern net-centric systemBy Haritha Treadway, Eurotech, Inc.Mil Tech Trends: Battlefield security –from the ground up34 Secure virtualization combines traditionaldesktop OSs and embedded RTOSs in militaryembedded systemsBy Robert Day, LynuxWorks38 Developing high-performance embeddednetwork security applications:A heterogeneous multicore processing approachBy Daniel Proch, NetronomeEVENTSensors Expo & ConferenceJune 7-9, 2010 • Rosemont, ILwww.sensorsexpo.comE-CASTwww.opensystemsmedia.com/eventshttp://ecast.opensystemsmedia.comStreamline Safety-Critical Software Development toStop Budget and Schedule OverrunsJune 8th • 1pm EDTPresented by: LDRA, Visure SolutionsWEB RESOURCESSubscribe to the magazine or E-letterLive industry news • Submit new productshttp://submit.opensystemsmedia.comWhite papers:Read: http://whitepapers.opensystemsmedia.comSubmit: http://submit.opensystemsmedia.com4 May 2010 Military EMBEDDED SYSTEMS

ADVERTISER INFORMATIONPageAdvertiser/Ad title21 Alphi Technology Corporation – The expert inmilitary technologies2 Annapolis Micro Systems, Inc. – Highperformance signal and data processing48 Curtiss-Wright Controls Embedded Computing– Speed your time-to-market3 D-TA Systems – 10 Gigabit sensor processing44 Elma Electronic – “The industry’s choice inVPX handles and panels“33 Elma Electronic – Systems – 10 GigabitEthernet switches blade level VPX12 Engineering Design Team Inc. – PCIe x8FPGA board13 Excalibur Systems, Inc. – Dragon47 GE Intelligent Platforms, Inc. – Welcome to theknowledge bank37 InnoDisk – Top reliability, finest solution7 Kontron – We do not build military aircraft23 LCR Electronics – Engineered for yourapplication41 LynuxWorks, Inc. – We won’t leave you highand dry45 Magma – Bridge telecom to legacy platforms26 Microhard Systems, Inc. – Wireless OEMsolutions18 Nallatech – High performance FPGA solutions42 Parvus Corporation – Qualified to perform5 Pentek, Inc. – Stroke of genius26 Phoenix International – Ruggedized 3U fibrechannel RAID system31 Schroff a Brand of Pentair ElectronicPackaging – Schroff understands thedemands12 TEWS Technologies LLC – COTS I/O solutions28 Themis Computer – Themis servers:speed to burn17 Trident Space & Defense – Protect your datawith Trident solid state drives25 Tri-M Systems Inc. – PC/104 Can-Tainer29 Tri-M Systems Inc. – 100Mhz PC/104 module10 White Electronic Designs – We create space19 Z Microsystems, Inc. – Rugged rackmounted serversE-LETTERwww.mil-embedded.com/eletterCloud computing offloads legacyhardware, balances loads, and putsthe focus back on the applicationInterview with CIA veteran Bob Floresand Appistry founder Bob LozanoData indexes boost embeddedsoftware’s performance andefficiencyBy Steve Graves andKonstantin Knizhnik, McObjectStatic analysis improvesefficiency, reduces downstreamintegration costsBy Andy Chou, CoverityMilitaryMilitary EMBEDDED SYSTEMSEMBEDDED SYSTEMSMilitary & Aerospace GroupChris Ciufo, Group Editorial Directorcciufo@opensystemsmedia.comSharon Schnakenburg-Hess,Assistant Managing Editorsschnakenburg@opensystemsmedia.comJennifer Hesse, Assistant Managing Editorjhesse@opensystemsmedia.comTerri Thorson, Senior Editor (columns)tthorson@opensystemsmedia.comMonique DeVoe, Web Content EditorSales GroupDennis Doyle, Senior Account Managerddoyle@opensystemsmedia.comTom Varcie, Senior Account Managertvarcie@opensystemsmedia.comRebecca Barker, Strategic Account Managerrbarker@opensystemsmedia.comAndrea Stabile,Advertising/Marketing Coordinatorastabile@opensystemsmedia.comChristine Long, Digital Content Managerclong@opensystemsmedia.comInternational SalesDan Aronovic, Account Manager – Israeldaronovic@opensystemsmedia.comSally Hsiao, Account Manager – Asiasally@aceforum.com.twEditorial/Business Office16626 E. Avenue of the Fountains, Ste. 203Fountain Hills, AZ 85268Tel: 480-967-5581 n Fax: 480-837-6466Website: www.opensystemsmedia.comPublishers: John Black, Michael Hopper,Wayne KristoffHermann Strass, European Representativehstrass@opensystemsmedia.comKonrad Witte, Senior Web DeveloperSteph Sweet, Creative DirectorJoann Toth, Senior DesignerDavid Diomede, Art DirectorPhyllis Thompson,Circulation/Office Managersubscriptions@opensystemsmedia.comRegional Sales ManagersBarbara Quinlan, Midwest/Southwestbquinlan@opensystemsmedia.comDenis Seger, Southern Californiadseger@opensystemsmedia.comSydele Starr, Northern Californiasstarr@opensystemsmedia.comRon Taylor, East Coast/Mid Atlanticrtaylor@opensystemsmedia.comReprints and PDFsNan Holliday800-259-0470republish@opensystemsmedia.comVice President Editorial: Rosemary KristoffVice President Marketing & Sales:Patrick Hopperphopper@opensystemsmedia.comBusiness Manager: Karen Layman6 May 2010 Military EMBEDDED SYSTEMS

Field IntelligenceBy Duncan YoungClustered GPUsclosing in onsupercomputerperformanceThe potential of high-end GraphicalProcessing Units (GPUs) to efficientlyprocess complex, real-time sensorimage and video data has been recognizedfor some time. The introductionof GPU-appropriate C-based computinglanguages such as OpenCL (KhronosGroup) and CUDA (NVIDIA) gives theprogrammer access to arrays of processingcores within each GPU, offeringtypically 500 GFLOPS of performanceto apply to highly parallel, data-intensivealgorithms. Combining multiple GPUsinto computing clusters – using industrystandard, open architecture formats –will provide TFLOP performancelevels for rugged, embedded sensorprocessing applications such as multisensor,multiplatform tactical battlefieldsurveillance.The tried-and-trusted solution for complexsensor processing is a heterogeneousprocessing configuration based on anFPGA front end, plus a multi computingarray of vector processors, such asFreescale’s 8640 Power Architecture.Using an FPGA front end is the idealsolution with its flexible high-speedI/O signals to interface to the sensor,plus it can be relatively easily appliedto repetitive, parallel algorithms such asfiltering and data reduction. For ultimatereal-estate efficiency, multiple ASICsor FPGAs have been used to replace vectorprocessors. However, this approachhas proven to be time consuming todevelop, test, and verify. Additionally,the promise of FPGA reconfigurabilityhas never truly been realized, makingoverall life-cycle costs and upgradeabilityprohibitively expensive for anybut the most generously funded projects.Much of this difficulty with portabilityand maintainability can be attributed toFPGA code being a specific hardwaredescription of the problem to be solved.This code often incorporates thirdpartycores and logic optimizations toachieve reliable operation over temperatureextremes.Software holds key to portabilityCUDA and OpenCL provide abstracted,multithreaded computing models, codedin C with extensions to suit a GPUspecifichardware architecture. As a result,the major manufacturers of GPUs, such asNVIDIA and ATI (AMD), do not need torelease detailed hardware descriptions ofthe processor-core arrays in each deviceas these will change as devices evolve.Currently, each processor core is a veryfast, single-precision math engine muchsimpler than a general-purpose processor.It has no register set as such, no predictivelogic, and no pre-fetch capability, butis designed to be constantly fed with dataand instructions. A typical GPU architectureuses multiples of these cores. (In thecase of NVIDIA, these are referred to asCUDA cores.)Eight cores form a group with commonregister sets and memory, known collectivelyas a streaming multiprocessor. Thisstreaming multiprocessor runs a simplekernel to dispatch operands and organizeresults from the cores. Each GPU hasmany such multiprocessor groups. Multithreadingis inherent within this type ofarchitecture as individual cores are onlyrequired to perform a math operation onwhatever data is presented to them, whichcan be from a different thread from cycleto cycle. Multithreading is transparentto the programmer, the compiler, anddevelopment tools dividing tasks amongstreaming multiprocessors to make themost efficient use of available resources.General-purpose host CPUOpenCL and CUDA require a generalpurposeCPU to host the application anddispatch tasks to a GPU. This is typicallybased on the desktop PC architecture,using the GPU as a math accelerator inaddition to its primary role as a graphicsdevice. This basic CPU/GPU configurationis mirrored by many off-the-shelf,embeddable PC products in formats suchas CompactPCI or VPX (VITA 46), someof which are available in ruggedizedform for deployment in harsh militaryapplications. However, by breaking theone-to-one relationship between CPUand GPU, even greater performance gainscan be made by creating clusters of twoor three GPUs accessible to just one CPUvia a high-speed PCI Express switch. TheVPX module format, in particular, providesthe capability to route PCI Expressoffboard, through the backplane, allowinga host CPU access to a cluster of GPUslocated on VPX modules within the samechassis. Such a configuration can yieldTFLOPS of performance and is embeddablewithin any deployable environment.Depicted in Figure 1, the NPN240from GE Intelligent Platforms adopts thisarchitecture to pack dual NVIDIA GT240GPUs and DDR3 local memory onto asingle rugged 6U VPX module.Figure 1 | The NPN240 multiprocessor fromGE Intelligent PlatformsOpenCL and CUDA offer unprecedentedlevels of performance for the repetitive,parallel processing tasks found inadvanced sensor systems. Importantly,these are software-based technologies,making them easier to develop, maintain,and port from one GPU to future generations.Looking ahead for rapid evolution,the next generations of GPU have alreadybeen announced, such as Fermi fromNVIDIA. Fermi will have many morestreaming multiprocessing groups, onchipcaches, 32 CUDA cores per group,more logic within each core, and full supportfor double precision floating-pointmath, suited to new high-performance,high-resolution sensor types.To learn more, e-mail Duncan atduncan_young1@sky.com.8 May 2010 Military EMBEDDED SYSTEMS

Mil Tech InsiderNetworked Attached Storageexpands data storage and securityon military platformsBy Steve EdwardsThe pervasive use of standard network fabrics such as Etherneton military platforms has joined with recent cost and densityimprovements in solid-state non-volatile memory, fueling significantgrowth in demand for Networked Attached Storage (NAS)devices the past couple of years. On today’s military platforms,on the ground and in the air, the use of Ethernet to connect variousparts of avionics architecture or crew stations, smart displays, andcontrols to the platform’s local network has become nearly ubiquitous.System designers have found that Ethernet enables them toeasily attach devices together over the network. Once networked,the platform’s subsystems can take advantage of the high-speed,high-volume data storage; Space, Weight, and Power (SWaP) mitigation;and enhanced security options that NAS makes possiblecompared to direct attached storage alternatives.NAS: Flexible storage in net-centric environmentsToday’s devices are incorporating network connectivity as systemdesigners embrace net-centric architectures. Legacy subsystemsbuilt with older technologies – such as RS-232, 1553, andCANbus – are being bridged to Ethernet so they can communicatewith other systems on the network. Now, using NAS, designerscan more fully exploit the Ethernet-based LAN to centralize datastorage that was formerly available only to individual devicesthrough direct attached devices. While direct attached storagedevices enabled system designers to dedicate private storage ofthe speed and size required by each particular subsystem, NASprovides the significant advantage of supporting easy and dynamicreallocation of available system memory. With NAS, if a newsystem requires storage capacity, the needed memory can be easilyallocated from existing memory without incurring the additionalcost, weight, and maintenance of a new storage device.NAS benefits: Security, version control, and SWaPAnother significant advantage of NAS over direct attached storageis that NAS enables easy removal of stored data, especially classifieddata, so that it can be protected remotely when the platformis not in operation. Direct attached stored data is located throughouta vehicle, making it difficult to remove for security purposes.With NAS, the user can remove individual cards or the entireNAS box. To further ensure data security, today’s NAS devicesalso incorporate encryption technology, such as 256-bit AdvancedEncryption Standard (AES) encryption. Vendors such asCurtiss-Wright Controls Electronic Systems (CWCEL) are alsoincorporating secure-erase technology to zeroize the encryptionkey on the NAS device that causes the data to become declassifiedas it can no longer be decrypted. This approach allows terabytesof data to be sanitized within milliseconds, versus the many hoursit could take to zeroize terabytes of solid-state memory data.NAS devices ease software configuration control because theentire platform’s processing elements boot from the NAS. Theuse of standard protocols enables all of the platform’s processingsystems to boot from a common storage device. The boot softwareand related binary programs can be loaded onto a single box,ensuring any system software updates occur simultaneously. Thiseliminates the risk of any single device getting updated whileothers are overlooked.The need for SWaP mitigation also makes NAS attractive tomilitary system designers. By packaging the solid-state storageinto rugged, high-bandwidth 3U VPX modules within the storagedevice, lightweight compact NAS devices can be built. CWCELoffers the Compact Network Storage (CNS), a very compact, half-ATR-wide NAS device. It deploys one-half TB of memory on3U VPX cards such as the VPX3-FSM (Flash Storage Module).The use of solid-state memory eliminates rotating disks’ susceptibilityto shock and vibration.Solid-state memory densities rise as costs dropIt is only recently, though, that the cost of solid-state memories hasmade this approach economically practical. The cost of 500 GB ofsolid-state memory has dropped from $1 million (with 80 MBpsthroughput) to $35,000 (with 800 MBps throughput) in just sevenyears. Furthermore, that same 500 GB once required a full ATRsizedbox, but today can be deployed on two 3U VPX-based cards.As memory density increases, it will be possible to deploy muchmore storage in the same footprint. For example, the CNS 0.5 TBis expected to increase to 1 TB by the end of the third quarter of2010. With the higher-density networked storage made possiblewith NAS, system designers can capture and store more information– such as streaming data – than was possible before.To the user, the NAS appears as a local drive on the system.Because NAS operates as a standard file server, it runs standardoperating systems and major network protocols. It facilitates networkarchitectures in military platforms in a way not possiblewith direct attached storage. It provides access to many networkstandards and communications mechanisms, bringing state-ofthe-artnetworking into vehicles. The NAS/file server, a serverdesigned to provide file services to clients on an Internet Protocol(IP) network, is depicted in Figure 1.Figure 1 | The NAS/file server is designed to provide file services to clients onan IP network.To learn more, e-mail Steve at Steve.Edwards@curtisswright.com.Military EMBEDDED SYSTEMS May 2010 9

We create space…White Electronic Designs’ lean microelectronic solutions increase boarddensity, solve component incompatibilities, reduce design complexity, andextend product life and environmental performance in military applications.We offer turnkey design, assembly and test of custom multi-chip solutions,and a wide range of standard military off-the-shelf space-saving solutions.Expand the possibilities; visit www.whiteedc.com/spacesaving.

... with more computing power per square inch.Our 4GByte SLC NAND flash SSD BGA provides big features in aminiaturized package, freeing up precious board space. It is engineeredspecifically for the defense market, designed to ensure data reliabilityin mission critical rugged and mobile systems. Constructed using a 32-bitRISC processor as its core storage controller, this SSD is also availablewith a hardware- and software-triggered security erase that meetsmilitary protocols.Learn more at www.whiteedc.com/ssd.602.437.1520 TEL 602.437.9120 FAXSecuring the Future Through the Power of DesignWWW.WHITEEDC.COM

dependencies, creating a new set of software dependencies. Dependingon the application, additional software dependencies are introducedsuch as a network stack, distribution middleware, databaseinterface, user interface, security services, and so on. A replacementor significant change in any of these areas greatly affects an application’sbusiness logic even if the software is well structured.Solution: Model Driven Architecture (MDA)One emerging solution for both previously stated problems isthe concept of the Object Management Group (OMG) ModelDriven Architecture (MDA). MDA defines three model levels:Computation Independent Model (CIM), Platform IndependentModel (PIM), and multiple Platform Specific Models (PSMs).The PIM models software application structure, behavior, andfunctionality using the Unified Modeling Language (UML),independent of the underlying platform technology. The PIM istransformed into a series of PSMs that targets specific areas ofthe underlying platform technology.One example of a PIM-to-PSM transformation uses a technologyknown as Executable UML, currently supported by a number oftool vendors. Executable UML translates a PIM into source code(considered a PSM by MDA). Executable UML utilizes a subsetof UML 2.x diagrams such as state charts, activity diagrams, classdiagrams, and sequence diagrams used to model the PIM. A codegenerator analyzes the diagrams and automatically generates asource code set based on the compiler that is attached to the tool.Some of the more advanced tools allow selection of the compiler(C, C++, Java, or Ada) and compiler vendor. The more advancedtools also provide capabilities of visual debug and automated testand are integrated with a configuration management environment.It has been shown that this technology can automatically generate75 to 80 percent of an application’s business logic from UMLdiagrams. The set of UML diagrams that captures the softwaredesign is now used to generate the source code set.Other PIM-to-PSM translations include UML 2.x ports, interfaces,and active objects. Ports are used to encapsulate interfaces amongapplication business logic and areas of the underlying platformtechnology (middleware, operating system, and so on). Port stereotypesare used to select PIM-to-PSM translation. For example, aport could be designated as a CORBA Publisher/Subscriber fordata distribution. Active objects are used to transform PIM objectsto PSM threads (or tasks) in the underlying operating system.MDA: Grist for software legacy architectureNew software technologies based on OMG MDA with currentlyavailable vendor tools offer tremendous potential to extend legacysoftware life expectancy. These technologies hold the promise ofsignificant improvements in ease of maintenance, rapid integrationof new capabilities, and the ability to migrate software tonew platform technologies. Sounds too good to be true, doesn’tit? Well, honestly, there is a price to be paid to achieve thesearchitectural improvements. Though far from its idyllic end state,today’s MDA capabilities still offer significant improvements tothe current state of legacy software migration.D.K. McKean is CTO at Advanced Fusion Technologies.He has more than 32 years of experience as a systemsengineer, software developer, and software architect, developingreal-time embedded and safety-critical software.He can be contacted at david.mckean@aft-worldwide.com.Military EMBEDDED SYSTEMS May 2010 13

Daily Briefing: News SnippetsBy Sharon Schnakenburg-Hess, Assistant Managing Editorwww.mil-embedded.com/dailybriefingUSJFCOM releases JOEThe United States Joint Forces Command (USJFCOM) recentlyreleased its Joint Operating Environment (JOE) 2010 report.(See www.jfcom.mil/newslink/storyarchive/2010/pa031510.htmlto download the report.) U.S. Marines Corps General J.N. Mattis,USJFCOM commander, says JOE “in no way constitutes U.S.government policy and must necessarily be speculative innature. It seeks to provide the Joint Force an intellectual foundationupon which we will construct the concepts to guide ourfuture force development” in creating upcoming U.S. policiesfor worldwide operations (Figure 1). Regarding technology, thereport predicts that key advances will continue at an explosivepace through 2030, including Electro-Magnetic Pulse (EMP)weapons, directed energy systems, laser systems, and HighPowered Microwave (HPM) weapons. HPM weapons attackelectrical systems, electronics, and ICs for ISR and commandand control, yet remain nonlethal and nonexplosive in urbanenvironments. Moreover, robotic systems, nanotechnology, andNanoenergetics (NE) – touted to “dramatically increase the powerand efficiency of explosives and propellants” – are also vital to theFriend or foe:The USAF will soon knowWhether in one’s social life or even in business ventures, it’simportant to know whether one’s associates are friends or foe.And it’s even more important in military endeavors, wherenational security is at stake. Accordingly, Raytheon Companyrecently delivered to the USAF the first incarnation of theIdentification Friend or Foe (IFF) equipment-compatibleKIV-77 Mode 4/5 crypto appliqué computer. Specifically, theKIV-77 computer provides combat-identification capability towarfighters engaging in surface, land, or air combat. KIV-77 isNational Security Agency (NSA) Type 1 certified, meaning thatit has authority to handle classified information. Meanwhile,Mode 4 signifies legacy applications, while Mode 5 designatesnext-gen data links, encrypted between transponders and interrogators,to decipher whether an approaching aircraft is friendor foe. The first KIV-77’s delivery was at least 60 days ahead ofthe contract deadline, Raytheon reports.2030 battlefield, JOE says.VME flies the (un)friendly skiesSome say VME is dead. However, the recent fulfillment of aUSAF $6.5 million B-1B bomber (Figure 2) upgrade order byGE Intelligent Platforms indicates VME is still flying high.The upgrade consisted of a Vertical Situation Display Upgrade(VSDU) based on GE’s 6U VME Octegra3 rugged graphics/videoprocessor and the designed-for-Octegra3 VIM2 rugged videoinput mezzanine. Since the “goods” were delivered early, primecontractor Boeing has a large window of time before VSDU’sflight testing, slated for early 2011. VSDU aims to enforce B-1Baircrew flight safety by providing: 1) a DO-178B certified BoardSupport Package (BSP), ensuring flight worthiness, thanks toFigure 1 | The United States Joint Forces Command (USJFCOM) recentlyreleased its Joint Operating Environment (JOE) 2010 report, which is “speculativein nature” yet provides “an intellectual foundation” for future Joint Forcebattlefield (and other) strategies.GE subcontractor Ultra Electronics Controls; and 2) “improvedprotection against hostile action” for the aircrew.Partnership serves the Italian NavyThe FREMM European frigate program, a joint effort of Italyand France, will soon benefit from a new partnership betweenZ Microsystems and Eurolink Systems. The agreement specifiesthat Z Microsystems will supply the Italy-based Eurolink Systemswith Z Microsystems’ ZX rugged computer servers series forthe FREMM program, which ultimately includes 10 frigates forthe Italian Navy and 17 for the French Navy. Eurolink will thenprovide Z Microsystems’ ZX1, ZX2, and ZX3 servers to anunidentified “global communication supplier,” who will integratethe MIL-STD-810G compliant servers into FREMM’s onboardcomputing systems. Z Microsystems’ logical nomenclaturedesignates its three “extended ATX form factor” wares as: ZX1 is1U high, ZX2 is 2U high, and ZX3 is 3U high.Figure 2 | B-1B Lancer, U.S. Air Force photo by Airman 1st Class Corey Hook14 May 2010 Military EMBEDDED SYSTEMS

Java-enabled combat system passesthe testModernization is the name of the game for many militaryprograms, and the USS Bunker Hill (CG 52) is a good case inpoint (Figure 3). Now suited with a Lockheed Martin-developedAegis Weapon System, CG 52 recently finished its full combatsystems’ operational trial. Key to Aegis’ operation is Atego’sAonix Java-based PERC Ultra virtual machine technology,which renders real-time, deterministic performance and providesvirtual machine management and instrumentation tools to fulfillmission-critical needs. Meanwhile, the Aegis Weapon Systemresides on 92 ships presently in military service worldwide – andis sported by maritime vessels owned by Japan, South Korea,Australia, Norway, and Spain.[Editor’s note: As Military Embedded Systems went to press, Ategoannounced the availability of PERC Ultra SMP, a variation of PERC Ultraused in the Aegis system but designed for multicore processor systems.Concurrent Java “garbage collection” doesn’t require stopping the entiresystem for routine Java housekeeping.]Figure 3 | USS Bunker Hill (CG 52), now sporting an Aegis Weapon System,recently finished its full combat systems’ operational trial.Samsung 3.4%Sharp 3.5%HTC 6.3%Figure 4 | Top 2009 smartphone vendors2009 Smartphone market shareApple14.9%18 others16.1%RIM19.4%Nokia36.4%More and more people aregetting “smart”Smartphones are of wide-ranging interest, both to the gadgetizedconsumer and the military commander, who thinks “if onlyI could check a sit rep from my iPhone.” Smartphones are alsoof interest to Forward Concepts, who recently published a470-page study of related trends, entitled “Smartphone Device &Chip Market Opportunities ’10”. The report indicates that 2009boasted an 18% growth in worldwide smartphone shipments,rising to $67 billion and 171 million units. Smartphone mobileInternet consumption also rose by 29% in 2009 as compared to2008. Forward Concepts forecasts that 2010 market share willfeature North America leading in smartphone consumption at22%, followed by Western Europe at 21.6% and China at 17%.Meanwhile, the top 2009 contender for smartphone market sharewas Nokia, followed by RIM, Apple, and others (Figure 4).A flurry of contracts appearsCurtiss-Wright (CW) has been busy lately, and is about to get evenbusier, thanks to three recently announced defense contracts:1) CW will provide an “approximately $25 million” AdvancedMission Management System (AMMS) to prime contractorNorthrop Grumman as part of the U.S. Navy’s Broad AreaMaritime Surveillance Unmanned Aircraft System (BAMS UAS).BAMS UAS proffers “persistent” ISR for identification, classification,detection, and tracking of littoral and maritime targets.CW will design and manufacture BAMS UAS AMMS units atits Santa Clarita, CA Motion Control facility with deliveriesstarting at the end of this year and continuing through next year.2) A second contract between Northrop Grumman and CW for$10.5 million stipulates that CW produces an upgraded RadarSignal Processing (RSP) ware for the Joint Surveillance andTarget Attack Radar System (Joint STARS) program. Work willbe performed at CW’s Motion Control divison in San Diego, CAand is one segment of an upgrade to the USAF’s E-8 JointSTARS aircraft’s Radar Airborne Signal Processor (RASP) radarsignal processing system. 3) And finally, a $1.17 million contractstipulates that CW provides its rugged, high-altitude capableSANbric Storage Area Network (SAN) to prime MDA inRichmond, BC, Canada. The SAN will then take flight on theCP-140, a Canadian Forces Air Command’s maritime patrolaircraft. CW’s Littleton, MA Electronic Systems facility willproduce the MDA SANbric units.For consideration in Daily Briefings, submit yourpress releases at http://submit.opensystemsmedia.com.Submission does not guarantee inclusion.Predator climbs to 1 millionMilitary tech trend reports indicate that Unmanned AircraftSystems (UASs) are increasingly being used on the modernbattlefield, with flexibility in configuration and the ability tospare human lives as the most critical factors. This technologyshift has been validated by a recent milestone: The Predatorseries of UASs recently reached 1 million flight hours. Trendstatistics leading up to the milestone include annual Predatorflight hours increasing from 80,000 hours in 2006 to 295,000last year. The Predator MQ-1B (Figure 5) surpassed 700,000flight hours this past March, with the remainder of the Predatorseries marking the other 300,000. Meanwhile, the Predatorseries includes the B/MQ-9 Reaper, C Avenger, Sky Warrior, andothers on active duty with the U.S. Department of HomelandSecurity, U.S. Army, USAF, U.S. Navy, NASA, the U.K.’s RoyalAir Force, and others.Figure 5 | The Predator MQ-1B, U.S. Air Force photo by Tech. Sgt.Sabrina JohnsonMilitary EMBEDDED SYSTEMS May 2010 15

Software: Nano-kernel commands fast deploymentPOSIX RTOS attacks SWaPand time-to-deployment issuesBy Kim RoweInfantry are limited by the Size, Weight, and Power (SWaP) capabilities of their equipment. The next generation ofhandheld and portable kits can exploit COTS Microcontroller Unit (MCU) technology to vastly reduce the cost of producingsmall, lightweight, ultra-low power gear. Adding an ultra-tiny POSIX RTOS enables the immediate introduction of scoresof standard applications and rapid adaptation of equipment to dynamic battlefield conditions, maximizing mission success.U.S. Army photo by Pfc. Sarah De BoiseChoosing a standard set of SoC MicrocontrollerUnit (MCU)-based hardwareand a standards-based RTOS opens thedoor to fast and flexible development oftools and sensors for use in the battlefield.Millions of lines of Linux code areavailable, and now they can be applied tohelp warfighters quickly and easily – andwithout significant SWaP requirements.This is a lean product developmentapproach that will minimize Total Cost ofOwnership (TCO).The key to reducing SWaP for portableelectronics is:• Minimizing total parts count• Building with die instead of packagedcomponents• Utilizing multichip modules• Exploiting consumer and automotiveparts• Exploiting IP• Utilizing a POSIX-based RTOSUsing this approach, existing applicationscan be quickly shrunk onto COTSSoC MCUs, saving money, time, power,and weight.How can a POSIX RTOS make such adifference to the hardware design? Byallowing the use of SoC MCUs to replacepower-hungry discrete implementations,electronics’ operating life can be increasedsignificantly while reducing size, weight,power, and cost. At the same time, thePOSIX compatibility allows rapid injectionof standard applications into devices,ensuring that software capabilities are notcompromised. This discussion exploreshow modern MCUs and power electronicsincrease SWaP, and where software andstandards-based RTOSs fit in. An ultra-tinyPOSIX RTOS example is also presented.Reducing power and weightincreases survivalA few years ago, MCUs were tiny eightbitmachines with very limited memorythat were clocked at a few MHz. Today,that has all changed. SoC MCUs runup to 200 MHz and have more than2 MB of flash, 128 KB of RAM, and allperipherals onboard. High-performancefloating-point hardware is available onSoC MCUs, speeding DSP applications,video processing, and automatic detectionalgorithms. The vastly increasedcomputing power of these devices in tinypackages allows battlefield computersto shrink from notebook sizes to handhelddevices with corresponding savingsin weight.The power consumption of MCUs hasalso dropped drastically. As little as oneyear ago, common thinking would havean 8-bit processor consuming less powerthan a 32-bit CPU. Today, using a 32-bitprocessor to replace the 8-bit processorand powering the processor down whenit is idle is more effective. Today’s 32-bitMCUs use a stunning 85 microwatts/MHz. On the battlefield, this meansincreased operational time, greatermobility, and less weight.Power electronics that enable MCUshave also shown significant improvement.Today Field Effect Transistors(FETs) run at 900 V, providing betterswitching characteristics at higher voltagesand reducing the power lost duringswitching compared to Insulated GateBipolar Transistors (IGBTs). Using softswitching techniques to minimize IGBTlosses (1/3 factor) makes this trade-off16 May 2010 Military EMBEDDED SYSTEMS

very application dependent. High-voltageFETs offer fewer components whileIGBTs offer smaller packages. Eitherapproach means less weight, improvedreliability, and greater mobility – increasingboth effectiveness and survival.Software and standards-basedRTOSs are keyPreviously, MCU software was 20 percentof the problem, and the main issueswere getting the hardware to work effectively.Now, 80 percent of the problem indeveloping an MCU system is software.The SoC MCU represents a completesystem on a chip with analog and digitalI/O, file systems and databases, networkconnectivity, and user interfaces. Furthercomplexity comes as these MCUs arenetworked together to solve larger problemsusing less power, with less weightand greater design reliability. A portablesolar converter, for example, would have1/6 of the power consumption, 1/10 ofthe weight, and greater than six times thereliability for the control electronics.The approach used to implement the softwarefor these systems is also changing.The four approaches that users typicallyselect are:• Single loop of control or timer-basedscheduler• Small scheduling kernel• RTOS (kernel plus I/O)• Standards-based RTOS (POSIX)The confusing part for designers is thateach of these approaches can be effective“Previously, MCUsoftware was 20 percentof the problem, andthe main issues weregetting the hardware towork effectively. Now,80 percent of the problemin developing an MCU”system is software.depending on how complex the problemis. By minimizing the TCO for the softwaredevelopment and maintenance overthe product line’s lifetime, substantialsavings are generated that are not readilyapparent to traditional engineers. Figure 1shows these four approaches and whenthey are most effectively used.If systems are trivial, such as those thatmonitor a single analog value and flag anout-of-range condition, the old-school singleloop of control is best. As complexitygrows, though, the advantages of having aseparate stack for each thread, thread priorities,standard APIs, and a standard I/Omodel grows. For complex applications, astandards-based RTOS is the only way toeffectively solve the problem. ProprietaryFigure 1 | A single loop of control or a simple scheduler was acceptable in the past; however, withincreased complexity, MCU systems require standards-based RTOS products with commercial support.Military EMBEDDED SYSTEMS May 2010 17

Software: Nano-kernel commands fast deploymentkernels require invention of I/O modelsand driver development while completeproprietary RTOS solutions eliminatesoftware reuse, cause training problems,slow field deployment, and can limit realtimeresponse.An alternative approach to a proprietaryRTOS is to utilize a POSIX-based RTOS.However, many of the benefits of a POSIXRTOS are hidden. Application portability,elimination of training, real-timeresponse, and reduced time to deploymentin theater are required to achievethe optimum savings over the life cycleof a product line. POSIX delivers thesevariables, while proprietary RTOSs andother standards do not.As developers of military systems, we mustbe cognizant of the immediate demandsof warfighters in the field by choosing aPOSIX platform application that can beimmediately and easily adapted to fieldedsystems. On the surface, adding a podcastfeature to a military radio using off-theshelfsoftware is a nonessential feature;however, if it is used to listen to messagesfrom families far away, its morale boostcould be substantial. Being able toquickly add features to systems using offthe-shelfsoftware is a key feature to meetnew battlefield demands.POSIX and Linux-compatibleRTOSs are optimalAs mentioned, use of a standards-basedRTOS is optimal for other than nontrivialsystems. By making the selected standardPOSIX, substantially more benefits areaccrued to the user by virtue of availableapplications, number of knowledgeableprogrammers, and training savings.The POSIX-based, Linux-compatible,32-bit ultra tiny Unison POSIX RTOS(Figure 2), for example, utilizes the samearchitectural design as its nano-kernel8/16-bit version: DSPnano.By moving to a nano-kernel architecture,the size of the system is kept very small.Unlike other Linux variants, the tinyPOSIX RTOS and its nano-kernel versioncan run in 2 KB RAM and 10 KBflash for an entire application includingI/O. As more features are used, the kernelgrows. The modularity of the nano-kernelarchitecture contributes substantially tothe reduction in size of the system. Inaddition, elimination of memory managementsupport, lightning fast interrupthandling, and embedded safety featuresmake the POSIX RTOS highly effectiveon SoC MCU architectures. With itstiny memory footprint and POSIX APIs,the POSIX RTOS allows many applicationsto run on tiny, low-power, andlow-weight SoC MCU systems where itwas not possible before.Thus, the complexity of the system issignificantly reduced using the nanokernelwith a layered POSIX and LinuxcompatibleI/O layer in comparison toany Linux alternatives. Any engineer canunderstand the basic kernel and I/O modelin a few minutes due to its modular architectureand isolated interfaces. Servers (ordrivers plus software glue) can be addedfor device specific uses quickly and easily.Additionally, if a specific algorithm hasa limitation in theater, adaptation can befast and easy.Flexible devices save livesUsing lean product development formilitary hardware based on SoC MCUtechnologies, new power technologies,and an ultra-tiny, POSIX-based RTOS or18 May 2010 Military EMBEDDED SYSTEMS

Figure 2 | The ultra-tiny Unison and nano-kernel DSPnano offer a nano-kernel architecture with anultra-tiny memory footprint, POSIX APIs, and a modular structure.nano-kernel RTOS increases flexibilityand reduces total cost of ownership. Thisbenefits the manufacturer and the military.Additionally, the ability to adapt an RTOSquickly in theater – coupled with loweredpower consumption, smaller sizes, and reducedweight – maximizes survivabilityfor the troops in the field.Kim Rowe is CTOand founder ofRoweBots ResearchInc. and has morethan 30 yearsof experience inbusiness managementand systemsengineering. He has been instrumentalin the startup of several companiesand several business units in thecomputer systems and servicesareas. He has extensive internationalexperience, having taken a broad setof software and hardware productsto market in more than 20 countries.He has published approximately35 papers and articles in variousjournals and magazines. Kim holdsboth an MBA from the University ofOttawa and an MEng from CarletonUniversity. He can be contacted atpkr@rowebots.comRoweBots Research Inc.519-208-0189www.rowebots.comMilitary EMBEDDED SYSTEMS May 2010 19

Special: FPGAs and IP for video processingFPGAs balance architecture, IP,power in electro-optical/infrared systemsBy Dr. Tibor Kozek, Juju Joyce, and Suhel DhananiModern Electro-Optical/Infrared (EO/IR) systems have become increasingly and exceedingly complex, and thereforedemand processing capabilities best offered by FPGAs. The most critical design challenge in these systems is that of combininghigh-performance sensor/video processing with low power consumption. To help solve this dilemma, key sensor-processing andvideo-processing algorithms – and how they can be implemented on FPGAs – are presented.The MQ-1 Predator carries the Multispectral Targeting System with AGM-114 Hellfire missile targeting capability and integrates electro-optical,infrared, laser designator and laser illuminator into a single sensor package. U.S. Air Force photo/Airman 1st Class Jonathan SteffenPortability and versatility, coupled with leading-edge COTStechnology, characterize many modern Aerospace and Defense(A&D) sensor platforms. Whether mounted on Unmanned AerialSystems (UASs), manpacks, or left behind as autonomous sensors,Electro-Optical/Infrared (EO/IR) systems have becomeexceedingly complex and demand processing capabilities bestoffered by FPGAs. Power constraints have also tightened as afunction of available energy and heat dissipation in low-footprintplatforms.These military imaging systems have become increasingly moresophisticated and incorporate multiple advanced sensors – rangingfrom thermal infrared, to visible spectrum, to even UV focalplanes. Not only do these sensor outputs have to all be corrected(defective pixel correction and color correction) and interpolated,but images from multiple sensors must be fused, overlaid, andfurther processed for local display and/or for transmission on thebattlefield. The key design challenge in these systems is to combinehigh-performance sensor/video processing with low powerconsumption. The following discussion focuses on some of thekey sensor-processing and video-processing algorithms and howthese can be implemented on FPGAs. And because an FPGAbaseddesign reduces component count while adding flexibility,system power is reduced.Typical sensor processing systemFPGAs are the platform of choice for almost all state-of-the-artEO/IR systems – since they meet the need for requirements forprogrammability, high-performance sensor/video processing,and low power consumption.20 May 2010 Military EMBEDDED SYSTEMS

In fact, each new generation of low-power FPGAs featuressignificantly lowers static and dynamic power consumption byutilizing a combination of architectural enhancements and lowercore voltages, coupled with geometric advantages resulting fromshrinking silicon feature sizes.Sensor processingOutput from image sensors used in EO/IR systems needscorrection by using algorithms such as those for non-uniformitycorrection and pixel replacement. While these algorithmstypically require only a few mathematical operations per pixel,calculations need to be done at the pixel rate and with data thatmight be different for every pixel. In this case, an FPGA is anideal platform because of the inherent parallelism in the architecture,as well as the ready availability of algorithm IP to realizevideo-processing functions.For non-uniformity correction, pixel-specific coefficients needto be streamed into the logic block implementing the correctionformulas. For smaller sensors, these coefficients can be storedinternally in FPGA memory. Depending upon the resolution ofthe sensor and precision of the coefficients, the memory requirementwill vary. For larger sensors, this data needs to be bufferedin external memory and read out in sync with the pixel streamfor every video frame. In either case, there is typically a needto change the corrective data set based on some selected parametersuch as the Focal Point Array (FPA) temperature that variesgreatly from ground to at-altitude.A typical data flow for such an algorithm would double-bufferthe correction coefficients to allow the relatively slow datastream from flash memory to complete before a new set of data isapplied. One of the key FPGA advantages is the ability to create adata pathway that fits the algorithm, rather than change the algorithmto fit a predefined architecture. This is critical for achievinglow power consumption. The massive I/O and large number oflogic elements in the FPGA allow easy implementation of parallelism,and off-the-shelf IP algorithms facilitate a straightforwarddesign. Some of the more typical sensor-processing algorithmsthat are available for FPGAs such as those in Altera’s Cyclonefamily are shown in Table 1.Sensor-processing algorithms for FPGAsDigital zoom / binningNoise filteringNon-uniformity correctionWide dynamic range processingLocal-area adaptive contrast enhancementPixel-level adaptive image fusionElectronic image stabilizationSuper-resolutionMotion detection and multi-target trackingTable 1 | Typical sensor processing algorithms available for FPGA devicessuch as Altera’s Cyclone family.Military EMBEDDED SYSTEMS May 2010 21

Special: FPGAs and IP for video processingAnother way that FPGAs reduce power in EO/IR sensor systemsis via a dramatically lower footprint. For example, Figure 1 showsan Altera Cyclone FPGA-based system that implements a thermalsensor with integrated processing. The FPGA performs real-timeimage enhancement, image stabilization, and digitally enhancedresolution – and can also drive an integrated microdisplay. TheFPGA power consumption in this case is ~500 mW.Existing ImageNearest NeighborBilinear scalingNew Image4 tapsFigure 2 | Different video scaling algorithms and their implementation witha video scaling IPFigure 2 depicts one such scaling IP core available for an AlteraCyclone FPGA. This function comes prebuilt with various“Lanczos” filter functions. The Lanczos multivariate interpolationmethod is used to compute new values for any digitallysampled data. When used for scaling digital images, the Lanczosfunction indicates which pixels and in which proportion in theoriginal image make up each pixel of the final image.Figure 1 | A thermal sensor system with integrated image processing.Video processing: Mixing and scalingMilitary EO/IR systems often include multiple image sensorswith outputs that must be fused together and displayed on acustom display with non-standard resolution. (“Non-standard”implies resolutions different from the typical desktop or laptopLCD.) A video processing system can be used to generate acomposite image from two video sources on a custom display.And a video data path inside an FPGA can generate a compositeimage from multiple sources.The input video is first formatted into the desired color space andis then subsequently scaled (resized) and mixed (alpha blended)with multiple other video streams. Scaling and mixing are amongthe most commonly used video functions, and they can be realizedusing off-the-shelf IP algorithms available for FPGAs.Scaling can be as simple as copying the previous pixel (or droppingit), or it can be implemented with complex interpolation filteringtechniques to generate a new pixel. Figure 2 shows the differencebetween the different algorithms that can be used for scaling.The graphic illustrates all the generated pixels (shown in solidblack) versus all the original pixels (shown in white). There aremany ways of generating the solid black pixels; for instance, thenearest neighbor algorithm copies the preceding pixel. A morecomplex way would be to take an average of the two neighboringpixels in both vertical and horizontal dimension. Sometimes thisis known as bilinear scaling – bilinear because it uses a pixelarray of size 2x2 to compute the value of a single pixel.Taking this concept further, one can compute the new pixel byusing “m” pixels in the horizontal dimension and “n” pixels inthe vertical dimension. Figure 2 additionally shows how a pixelis generated using four pixels in each dimension – also called afour-tap scaling engine.Of course, the trick deals with the weights assigned to each pixel– also called the coefficients when realized algorithmically. Thecoefficients will determine the quality of the scaled image.Selecting from a range of Lanczos algorithms to scale the imageor bypassing them completely in favor of custom coefficients isalso possible. In either case, the function automates the tediousjob of generating HDL code for what is essentially a twodimensionalfilter. It also maps it to the various FPGA structuressuch as the DSP blocks and the embedded memory blocks, thusimproving productivity and reducing overall design time.Video overlaysThe other commonly utilized function is mixing and overlay oftwo or more video streams. This is generally done by an alphablending function. This is a way of generating a composite pixelfrom two or more pixels. One pixel is assigned an opacity valuecalled the alpha. When alpha is zero, that pixel is completelytransparent (that is: not displayed). When that same alpha valueis 1, the pixel is completely opaque – only that pixel is seen andthe other pixel is not displayed.In mathematical terms, the value of the composite pixel iscalculated as:C = αP1 + (1-α)P2Whereα is the alpha valueP1 is the pixel 1 from the video layer 1P2 is the pixel 2 from the video layer 2C is the composite pixelThe same technique can be used to create translucent imagesbecause the alpha value can be set anywhere between 0 and 1.A more sophisticated way of combining information from two(or more) images is to utilize image-fusion algorithms. Imaginea thermal infrared sensor and a visible image sensor depictingthe same scene, but each contains information in different partsof the image.If alpha blending could be applied to select how dominant one orthe other image should be in the combined output, there wouldbe no single value for alpha that would extract all the informationavailable from the sensors. Figure 3 shows such a scenario.22 May 2010 Military EMBEDDED SYSTEMS

While the visible image on the top left contains informationabout the surroundings, the thermal view on the right only showsdiscernible features where there is a temperature differenceagainst the background. Conversely, the performance of the thermalsensor is unaffected by the strong light source in the scene,while the visible camera provides no information in the samearea as a result of saturation.In the fused view, a seamless combination of information fromboth input modalities is achieved on a pixel-by-pixel basis.Details missing from one modality are “filled in” from the otherand vice versa. One of the simplest forms of fusion would be toapply alpha blending with a different alpha value for every pixelcalculated from local image statistics. State-of-the-art fusionalgorithms, however, typically go beyond that and perform adecomposition of the input images that extracts relevant featuresaround every pixel. These features are then combined to form afused image.Dr. Tibor Kozek is cofounder/Chief Technology Officerof Imagize LLC and has nearly 20 years of experience insignal and image processing. Prior to Imagize, he was chiefscientist with Teraops, Inc., as well as a visiting scholarat the University of California, Berkeley.Juju Joyce is Senior Strategic Marketing Engineer,Military & Aerospace Business Unit, at Altera Corporation andhas more than 10 years of semiconductor industry experience.He holds a Bachelor’s degree in Electrical Engineering fromthe University of Texas at Austin.Suhel Dhanani is Senior Manager, DSP, for AlteraCorporation’s software, embedded, and DSP marketing group.He has more than 15 years of industry experience with Xilinx,VLSI Technology, Anadigm, and Tabula. He holds M.S.E.E.and M.B.A. degrees from Arizona State University.Figure 3 | Sensor fusion allows simultaneous utilization of multipleEO sensorsAltera Corporation408-544-7000newsroom@altera.com • www.altera.comMilitary EMBEDDED SYSTEMS May 2010 23

Military Materiel: IT and the GIG – Front and centerTelecom equipment netsbig benefits on the battlefieldQ&A with Anthony Ambrose, VP and GM of RadiSysEDITOR’S NOTEEditor Chris Ciufo jumped at the chance to learn how RadiSys, one of the larger COTS telecom companies serving theembedded space, is finding homes for its technology in Aerospace and Defense (A&D). RadiSys is being pulled into themarket by DoD platforms demanding network-centric equipment in semi-rugged deployments. Edited excerpts follow.MIL EMBEDDED: So can you start byfamiliarizing our readers with RadiSys?AMBROSE: Sure. RadiSys has been inbusiness for 22 years. For the first 12 to15 years of our life, the customer wouldsketch what they wanted on a napkin, andwe would design and deliver it to greatresults. And, then really after the telecombubble of 2001, the whole dynamics ofthe industry changed, and not just in telecom.So it was pretty clear that we neededto go to more of a standard embeddedproduct portfolio, targeted to customerswho cannot have the productsfail.MIL EMBEDDED: Wheredoes the military focuscome in?AMBROSE: About a yearor 18 months ago, we said,“We’re not really targetingthe military space, so whyare military customerschoosing our products?”We talked to some experts in the industry,and they came back and said, “Look,RadiSys, you’ve got some great productshere. You’ve got about 90 percentof what the military and aerospace marketalready needs. You just have to makea few simple changes in some of yourbasic processes and basic marketing,highlight some of the things that you’realready doing, and you’ll have a verygood mil/aero solution including ATCAand COM Express.”And it turns out that for applications likeground stations or command and control– or anything needing extremely24 May 2010 Military EMBEDDED SYSTEMShigh compute performance – ATCAis a superior solution, far in excess ofwhat VME or CompactPCI or even VPXwould deliver. And ATCA has been onthe market for six or seven years so wehave learned a lot and earned a high levelof trust from our customers. Our ATCAproducts are very much field-proven.MIL EMBEDDED: Skeptics mightsay, “Those comm guys don’t speakour language, and we don’t understandthose telecom standards likeMicroTCA or ATCA.”... For applications like groundstations or command and control – or anythingneeding extremely high compute performance –ATCA is a superior solution, far in excessof what VME or CompactPCI or evenVPX would deliver.AMBROSE: You know, I’d be surprised ifmany were saying that. The reason is thatATCA has adopted almost the exact oppositeapproach of MicroTCA. And asidefrom having three letters in common,there’s really not much commonalitybetween the two. ATCA is field provenand deployed, and it’s been discoveredby customers as very useful for the militaryand aerospace markets. ATCA was“stealth” technology, meaning it had nomarketing energy behind it in the militaryand aerospace market. So it’s beenselected with no hype, with no buzz, withno top-down standard because it’s thebest product, period.MIL EMBEDDED: So will we beseeing more of RadiSys targeting themilitary market?AMBROSE: Yes, we’re absolutely targetingmore for military and aerospacemarkets. It’s a pretty straightforwardstrategy. We’ve selected products thathave traction in these segments already,such as ATCA for C4ISR and informationassurance and COM Express forr uggedized computers, robots, UAVs,and UGVs.MIL EMBEDDED: Whichtechnologies are youbringing to the militaryarena that perhapsweren’t there before,given your telecombackground?AMBROSE: We’ve had anopportunity to work withseveral of our technologypartners to develop a ruggedizedchassis, and we’vecome up with a conformal coating capabilityfor ATCA. So now we can provide aruggedized ATCA chassis, which expandsthe market even further.MIL EMBEDDED: I’ve not heardof a ruggedized ATCA chassis. Solet’s talk boards. A typical benigncommercial temp range wouldbe 0 to +50 °C, or maybe 0 to +70 °C.Are your boards rated for a widertemperature range than that?AMBROSE: Yes, we deliver a wider temperaturerange than that, specifically on theCOM Express. [Editor’s note: RadiSys’

Intel Atom-based Procelerant CEZ5XTand CEGSXT COM Express modules bothboast an operating temp range of -25 °Cto +70 °C.] Another point to bring up isthat a lot of people use a simple screenwhen testing; that is, they test a productone time in the factory then if that happensto work at that lower temperature,they certify it as having some extendedtemperature properties. However, weutilize both HALT and HASS test methodologies.This ensures that a productworks to the extreme of its temperaturerange; unlike other products certifiedbased on a single test, it doesn’t leave thefactory until we see those results.MIL EMBEDDED: You’ve describedyour company’s ATCA products severaltimes as “field proven,” which mightbe analogous to “mission critical.” Butisn’t an ATCA board just a commodity?AMBROSE: Well, I would characterizeATCA as an open standard, not a commodity.But even working with an openstandard doesn’t mean that everyone haschosen to invest at the same level, or willactually have the same experience level,or will build the same types of products atthe same quality level. So, again, RadiSysfocused on more of a premium positionin the market, where our products gointo applications that just can’t fail. By“field-proven,” we mean our productshave been successfully deployed in theseapplications.In fact, we’ve had our systems NEBScertified, which is very important in telecom.A lot of this is very similar to whatyou need in military and aerospace: Shockand vibe tests, for example in NEBS, electrostaticdischarge, flame tests, things likethat … the ability to work in a very hotenvironment. When you certify to NEBS,your product has to work at up to +55 °C.MIL EMBEDDED: Why do you thinkyour military customers are turning totelecom-based technologies?AMBROSE: We’re seeing a new set ofapplications emerge, outside of the standardavionics/vetronics space. It doesn’tmean customers don’t want to haverugged, reliable systems, but it meansthat they’re in a network-centric warfaremode. So it’s very logical that they wouldwant to have solutions that look morecommunications-centric.MIL EMBEDDED: OK, so they’rewilling to use a different set of metricsto describe and procure thoseproducts then?AMBROSE: Right, they still want rugged,they still want field proven, but in anumber of cases, we’re selling a COTStechnology. Because we’re selling a commercialtechnology, they know it’s open.When they choose ATCA, they want toknow it’s not something where RadiSysor anyone else can suddenly say, “Wait aminute, I own this, you can never bid itfrom anyone else in the future.”MIL EMBEDDED: What about longlifesupport for your ATCA products?Telecom isn’t known for lengthy EOLmanagement practices.AMBROSE: Actually, telecom equipmentdoes need to survive long-term. Becauseof our history in embedded segments thatdemand long life, we already know howto provide long-life solutions for typicalmilitary and aerospace requirements. Andwe’ve done that. We can point to customersthat say, “Look, here’s a product wedesigned in 1995 and we supported itall the way to 2007. Here’s a product wedesigned in 2001 and it’s still going.”MIL EMBEDDED: Whom would youconsider your biggest competitors inthe market?AMBROSE: The biggest competitoroverall across all of our businesses continuesto be in-house design: The conceptis a lot of people “roll their own boards,”and we understand why. They used todo that in commercial, they used to dothat in communications. Those marketsnow are rapidly adopting COTS technologythat allows them to deploy theirR&D resources to areas of their owndifferentiation.Then you have the traditional CompactPCIand VME companies. And we’re nottargeting direct replacement of those formfactors; but I certainly think that they’regoing to want to take their products andtry to extend them as far as they can. Ourcompetitors are also the PC/104 [smallMilitary EMBEDDED SYSTEMS May 2010 25

Military Materiel: IT and the GIG – Front and centerform factor] companies that could potentiallydo something with PCs. But realistically,the COM Express way of buildingrobots, UAVs, and ruggedized personalcomputers is a very robust method.MIL EMBEDDED: So far we’vefocused on boards; let’s shift gearsand talk about systems. Does RadiSysdo systems integration?AMBROSE: Yes, we’ve been doing that fora while. We’re going to see more and morecustomers say at the end of the day, “Justplease manage the whole thing for me.”MIL EMBEDDED: Which technologiesdo you foresee into the future?AMBROSE: ATCA is a very rich groundfor technology. We’ve just announcedATCA 4.0 – our first 40 G products andstrategies. You can fast-forward two orthree years and this will become veryimportant across a broad range of applicationsthat need such high performance.And that means if you have backplaneperformance, you can get a lot more I/O,even if your base technology’s around10 G.I also think the whole concept of integratingpacket processing and applicationsprocessing for secure networks informationassurance is gaining traction. Thatmeans you can do deep packet inspectionon a lot of your communications,and you can improve security withinyour networking infrastructure. It alsogives you the ability to have a lot of performanceso that even if you’re securingand encrypting information within thenetwork, you still get a pretty reasonableuser experience on either end.MIL EMBEDDED: What would you sayto those who remain skeptical aboutusing telecom-based technologies suchas ATCA in the military?AMBROSE: Unlike some other technologiesthat have been out there, as Imentioned, the marketing has lagged thereality in this case. We know [ATCA]products work, we know the products fita need in the mil/aero industry, and we’regoing forward on that basis.Anthony Ambrose joined RadiSysin 2007 and is the Vice President andGeneral Manager of CommunicationsNetworks. Previously, Anthony wasIntel’s General Manager of theModular Communications PlatformDivision in the CommunicationsI nfrastructure Group. In this role, heled Intel’s ATCA effort while hisorganization had the responsibilityfor servers, blades, boards, andsoftware for the telecommunicationsindustry. Anthony holds a B.S.in Engineering from PrincetonUniversity. He can be contacted atanthony.ambrose@radisys.com.RadiSys503-615-1100www.radisys.com26 May 2010 Military EMBEDDED SYSTEMS

Military Materiel: IT and the GIG – Front and centerDesigners are demanding end-to-endsensor processing, right out of the boxAn interview with Dipak Roy, Chairman of D-TA SystemsEDITOR’S NOTESure, there are lots of signal processing, I/O, and FPGA co-processor boards on the market. And many of them areamazingly fast, sporting leading-edge performance and software. The trouble is, designers still need to design a systemaround these boards, which often come from disparate suppliers. I caught up with Dipak Roy, an old friend from my COTSsonar days, who has resurfaced with his new company, D-TA Systems. Dipak is a bona fide sonar expert, and his formercompany, ICS (aquired by Radstone, now a part of GE Intelligent Platforms), invented the FPDP interface that became aVITA standard. The fact that Dipak is now relying on 10 GbE for data transfer came as no surprise. Edited excerpts follow.MIL EMBEDDED: According to yourliterature, D-TA’s value propositionis box-level, preconfigured systems.What does that mean?ROY: The conventional norm for buildingembedded systems for military applicationsis to start from COTS boards fromvarious vendors. One has to select a computerplatform, OS, enclosure, and powersupply, then procure software devicedrivers and application routines.Next, the processesof system integration anddebugging commence, followedby laboratory testsand field trials. The wholeexercise is time consuming,expensive, and oftenleads to significant cost andtime overruns.However, D-TA Systems’approach changes thewhole paradigm for demandingsensor-processing applicationsby offering box-level, user-reconfigurablesystems based on 10 Gigabit networktechnology. Since our products come fullytested and require minimum customerreconfiguration, the deployment cost andtime savings are dramatic.MIL EMBEDDED: Describe the typesof integration challenges customersface with board-level point solutions.multichannel system is a time consumingand expensive proposition. For example,a user trying to implement a 16-channelHF radar system with four-channel COTSPMC ADC cards (the maximum ADCcount available on current PMC cards)needs to integrate four such cards, acquirea clock source, and develop a clock distributionstrategy to maintain synchronizationbefore any application developmentcan commence.... I believe that the current trendin multithreaded software processing inplatforms with multiple GPP coreswill revolutionize sensor processing; this will allowmore real-time signal processing functionsto be housed in software and GPPs,rather than in FPGAs.To help solve these issues, a fully integratedand tested functional module thatcan be quickly configured for a particularapplication is a real time-saver. Followinga successful trial, customers often requirecustomization services for repackaging,ruggedization, functional modifications,and so on, for volume deployment.MIL EMBEDDED: What are your targetmarkets and applications?sensor-array processing, with “plug-andplay”products that cover an applicationspace from sonar/acoustic to radio/radar.We are targeting a very wide range, fromsonar to wireless. In the sonar/acousticarea, we are looking at high-frequencyand high-channel-count applicationssuch as mine sonar, shallow-water sonar,obstacle avoidance, vibration analysis,ultrasound, and real-time acoustic simulation.In the radio andradar area, the productsare particularly suited tomultiantenna applicationssuch as phased array radar,RF direction finding, smartantenna base station, signalintelligence and target location,sonobuoy, RF test,and channel simulation/emulation. Our specialtyis to offer precise synchronization,from RF tobaseband, across a largenumber of input and output channels.Interestingly, we have also recently establishedan advanced sensor processinglaboratory at a local university, andseveral faculty members and graduatestudents are now engaged in developingnew and exciting applications.MIL EMBEDDED: Where does 10 GbE fitinto the need for deployment speed?ROY: Starting from board-level products(COTS boards) and building a complexROY: Defense, aerospace, wireless, andtest and measurement. Our specialty isROY: 10 Gigabit network technologyoffers data throughput rates that are fasterMilitary EMBEDDED SYSTEMS May 2010 27

Military Materiel: IT and the GIG – Front and centerthan any computer bus. And, more importantly, it allows better system partitioning thatis particularly advantageous in mixed-signal applications. Analog functions can be isolatedfrom the “noisy” computer bus for better analog performance. Moreover, it is partof the IEEE standards process and is continually upgraded for speed and bandwidth.MIL EMBEDDED: You are building systems using the same COTS hardware andsoftware as everyone else. Why would a customer choose your solution?ROY: All our products are designed from the ground up. Unlike the typical COTS boardvendor, we are not constrained by board size, predefined connector pin-outs, powersupply ratings, enclosure size, and so on. So we can offer an uncompromising solutionthat is designed to solve the problems encountered in demanding sensor-processingapplications: dynamic range, analog performance, processing speed, and instantaneousbandwidth, to name a few.Themis’ New Rugged Servers HaveSpeed to Burn and Keep Their Cool.New! 1RU RES Servers- One or two Intel ® Quad-Core 5500 Series Xeon ®CPUs with Intel Nehalem Microarchitecture- Up to 96GB ECC SDRAM- Up to 3 removable and lockable 2.5” HDDs- One PCI-E 2.0 x16 slot, optionalSAS expansion- 2RU RES Servers also availableRES-12XR3 server shown with optional filter door panels.RES-32XR3 server shown with optional filter door panels open.New! 3RU RES Servers- One or two Intel Quad-Core 5500Series Xeon CPUs with IntelNehalem Microarchitecture- Up to 144GB ECC SDRAM- Up to 8 removable and lockable2.5” HDDs- Up to 7 expansion slots(PCI-E and PCI-X)Transformational.A New Eraof Performanceand Rugged ReliabilityThemis’ new family of XR3 Seriesof Rugged Enterprise Servers (RES)includes the latest Quad-Core Xeonprocessors and Nahelam Microarchitecturefrom Intel. These new Intel chips revolutionizeserver performance, and Themis’ robustdesigns – only 20” depth - provide thereliability to keep mission critical applicationsrunning. Themis servers provide far greaterreliability, improved life cycle management andsubstantially lower TCO than other COTS systemssolutions.Features in the RES-XR3 servers include:- Dual redundant, hot-swappable power supplies- Dual redundant DC power option- Operating shock - 3 axis, 25G, 20ms- Operating vibration - 3.0 Grms, 8Hz - 2000Hz- Light weight, corrosion resistant, 20” depth chassis- Optional air filter door panelsSo when the environment gets tough and your datais critical, turn to the company that builds systemsto perform in the harshest conditions. For Sun®Solaris, Linux®, and Microsoft® Windows®environments. For more information on Themis’rugged new servers, please visit www.themis.com.Themis rugged, mission-critical computers.Designed to take it.(510) 252-0870.©2009. Themis Computer, Themis, the Themis logo, and Rugged Enterprise Servers are trademarksor registered trademarks of Themis Computer. All other trademarks are the property of their respective owners.Also, as mentioned, virtually all of ourproducts are 10 Gigabit network attached,and most of them are housed in a 1U high,standard 19" rack-mountable enclosure.Our mandate is to offer end-to-end solutionsfor demanding sensor-processingapplications. For radio and radar applications,examples include tunable RF,multichannel IF (software radio), and10 Gigabit record and playback systems.For high-frequency sonar and acousticapplications, there are high-precisionsignal conditioning, 24-bit digitization,and a high channel count. These productscan be seamlessly connected to build anycomplex sensor processing systems in amatter of days, not years. Other factorsin making a designer’s life easier includelibraries of pretested FPGA cores andreal-time multithreaded software applicationroutines. So our value proposition isvery compelling: The customer can focuson their applications and not on dataacquisition system development.MIL EMBEDDED: What are the softwareconsiderations?ROY: The 10 Gigabit network connectionis OS agnostic, so we typically use the10 Gigabit links for data transfer and1 Gigabit network for control. The baseSDK supplied with the products includesa Control API and Data API. The ControlAPI allows the control of our boxes, whilethe Data API allows the user to build applicationsusing example codes provided.This structure shields users from socketcalls and simplifies data access. We havealso developed many DSP functions formultithreaded, multicore platforms thatseamlessly attach to the base SDK.Looking toward the future, I believethat the current trend in multithreadedsoftware processing in platforms withmultiple GPP cores will revolutionizesensor processing; this will allow morereal-time signal processing functions tobe housed in software and GPPs, ratherthan in FPGAs. It will also save developmenttime and cost and enable rapidreconfiguration.MIL EMBEDDED: Will the sensorprocessingworld adopt emerging 40 oreven 100 GbE technology?ROY: D-TA Systems is the first company tointroduce 10 Gigabit sensor processing. Itis new and just getting started. As opposed28 May 2010 Military EMBEDDED SYSTEMS

to datacom applications, sensor-processing applications require sustained processing at ahigh data rate. Most applications cannot tolerate missing samples or data packets. Thereis no “resend” in sensor processing applications. The processor has to keep up with thedata rate. That is why the sensor-processing world trails behind the datacom world interms of sheer network speed.As the processor gets faster and faster, we will see introduction of 40 and 100 GbEnetworks for sensor processing applications for the simple reason that more signal channelswith higher bandwidths can be handled by each network. Moore’s law dictates thatthe processing power will increase to allow signal processing at these rates.MIL EMBEDDED: How has sensor-processing technology evolved in the past20 years?ROY: One has to remember that not long ago, all sensor processing was done in the analogdomain. Over the years, it has evolved significantly and in many ways. Today, the processinghas become mostly digital, primarily because of increased data converter speedand accuracy. This has not only allowed detection and processing of low-level signals,it has also allowed data converters to get closer and closer to the actual sensor, therebysimplifying front-end analog design. For example, the advent of Sigma-Delta technologyhas revolutionized acoustic signal processing. The advent of FPGAs has allowed manyDSP functions to be implemented in real time. The availability of high-speed networksand computer buses has allowed raw or preprocessed sensor data to be accessed by ageneral-purpose computer for further processing or display.MIL EMBEDDED: What are the biggest challenges facing your customers?ROY: The biggest challenge facing our customers is time to market. In the current tightbudgetscenario, cost overruns cannot be tolerated. Further, most customers these dayswant to see a working demonstration before funding any development. D-TA products aredesigned to precisely address these problems: Virtually any sensor-processing systemscan be built quickly by selecting products from our catalog.MIL EMBEDDED: Which new technology do you think we’ll see in the next two yearsthat’s not evident today?ROY: Going forward, we will see more evolution in data converter technology in terms ofhigher speed and accuracy. The other focus will be on lower-power devices that will alsoallow increased channel counts. During the next several years, it might be possible toconnect the data converters (ADCs and DACs) directly to the antenna. This will requirenot only very high-speed converters but also faster FPGAs and processors to handle thevery high data rate. In the short-term, we will also witness significant evolution in RFIC[Radio Frequency Integrated Circuit] technology, which will significantly reduce the sizeand cost of RF subsystems.Dr. Dipak Roy is Chairman of the Board at D-TA Systems Inc. His embeddedindustry experience began briefly in the form of Research Engineer. Then he tookthe entrepreneur route: His first company, ICS (now part of GE Intelligent Platforms),won defense contracts including the U.S. Navy’s SQQ-89 sonar upgrade program.Additionally, Dipak has authored a patent and published articles in 30 technicalpublications. He has received many awards, notable among them is the AmericanNational Standards Institute (ANSI) award for the invention and standardization ofFPDP, a high-speed data flow concept that simplifies system integration. In 2007, hewas appointed by the government of Canada to the board of Sustainable DevelopmentTechnology Canada (SDTC), a $1 billion foundation developing “Clean Technology”companies in Canada. He received his PhD in Electronics Engineering fromCarleton University, Ottawa. Dipak can be contacted at dipak@d-ta.com.D-TA Systems Inc.613-745-8713www.d-ta.comMilitary EMBEDDED SYSTEMS May 2010 29

Military Materiel: IT and the GIG – Front and centerElements of a deployed,modern net-centric systemBy Haritha TreadwayIn striving for the U.S. DoD’s vision of net-centricity, military embedded designs must optimize size and power consumption,provide fast and effective graphics and visualization, keep commanders and soldiers connected, and provide hardwaresupported by the Global Information Grid’s (GIG’s) Service Oriented Architecture (SOA) framework. But don’t forget thekey ingredient: the processor, which heightens performance and brings all these elements together.U.S. Marine Corps photo by Cpl. Albert F. HuntFor the past 10 years, the United StatesDepartment of Defense has pioneeredthe military doctrine of net-centricity toachieve effective information sharing ina complex environment. Net-centric warfareaims to achieve a robustly networkedforce for shared situational awareness,which in turn dramatically increasesmilitary mission effectiveness.Net-centric devices include soldiernavigation tools, precision optics capabilities,and radio technologies. These devicesrely on robust embedded computers andhigh-performance, low-power processors.Developing embedded systems for netcentricdevices includes the ability toreduce soldier-borne weight, increasemission length, and improve soldiersituational awareness and effectiveness.To achieve these goals, embedded designshave to optimize size and power consumptionwithout making sacrifices in equallyimportant graphics, I/O, and communicationscapabilities. In addition, softwareplays a key role in allowing embeddeddesigns to seamlessly and cost-effectivelyintegrate the hardware capabilities withexisting infrastructures developed to supportnet-centricity. Meanwhile, the underlyingprocessor accelerates performance andties together all these net-centric needs.Size and powerSize and power considerations are centralto the effectiveness of net-centric devices.Net-centric warfare leverages the powerof real-time information transmitted overa complex grid of data to make militaryoperations more effective over land,air, and sea. With instant access to datacomes the need for soldiers to be mobileand responsive, meaning the embeddeddevices they operate are flexible, compact,and light, allowing them to move freelyand quickly.In addition to being physically worn orcarried on a soldier, embedded devicesalso operate in vehicles such as UAVs,specialized planes, and helicopters thatare space-constrained while still requiringthe ability to interface with extensivemachinery and other peripherals. As aresult, size is one of the top priorities indesigning embedded systems targeted fornet-centric warfare.By choosing processors with low ThermalDesign Power (TDP), designers can30 May 2010 Military EMBEDDED SYSTEMS

“3D visualizationand simulations withinteractivity are criticalin understanding localenvironments in greatdetail – not only for theeffectiveness of themission, but also for thesafety of soldiers.”devices is a CPU with an integratedgraphics engine to allow for small formfactor designs.Connectivity and I/OWhile size, power, and graphics arecritical, another key to net-centric warfareis the communications and I/O capabilityApplication Soldier in field Control center SimulationsGraphicalrequirementsSmall screensVGA resolutionLightweightLarge displaysXGA/HD resolutionFast response timeof the device. Net-centric warfare reliesheavily on sensors, satellite communications,and GPS. These functions will belinked together via the IP-enabled GlobalInformation Grid (GIG), a communicationsproject of the U.S. DoD that connectsmilitary entities and operations acrossthe globe to a common networkedTable 1 | Graphical needs for net-centric warfare depend on the end application.High definitionXGA/HD resolution2D/3D graphics and fast response timeremove the need for bulky fans or heatsinks, thereby minimizing the size ofthe hardware. In addition, becausenet-centric devices are so mobile, powerconsumption is equally important, witha focus on battery operation easily lastingmore than 10 hours. To meet theseneeds, processors with features such asDynamic Voltage and Frequency Scaling(DVFS), deep sleep, and idle states bringan obvious advantage by managing heatand power dissipation. The challenge is tomeet these size and power consumptionneeds without sacrificing processing andperformance capabilities.Graphics and real-timevisualizationAs military operations become quickerand more reactive, net-centric warfarerelies heavily on sophisticated graphicsto provide realistic imagery of situations.3D visualization and simulations withinteractivity are critical in understandinglocal environments in great detail – notonly for the effectiveness of the mission,but also for the safety of soldiers.Display support is also varying, as shownin Table 1. On one hand, real-time simulationdisplays must be large with very highresolutions, in the range of XGA (1,024 x768) to High Definition (1,280 x 720). Fordevices carried by soldiers in the field, thedisplays are often smaller with VGA resolution(640 x 480). Dual-display supportis also important. Therefore, the key is topick an embedded processor that has theright mix of graphics capabilities for thespecific end application. An ideal choicegiven the space constraints of embeddedMilitary EMBEDDED SYSTEMS May 2010 31

Military Materiel: IT and the GIG – Front and centerinfrastructure. Figure 1 highlights howthe GIG offers an overarching structurefor global, cross-platform networking.IP-enabled Global Information GridFigure 1 | IP is used to connect variouscommunication devices (Ethernet, GPS, satellite,and radio) to a common infrastructure via theGlobal Information Grid.An SOA framework separates the hardwarefrom the end service or applicationlogic. SOA bundles are more portable andaccessible, easily transported from oneplatform to another with minimal developmentwork. Portability is critical innet-centric warfare, especially as projectslike the GIG require devices to be flexibleacross terrains and operations. Figure 2shows an example framework based onJava. Here, the advantage is to separateapplication-specific bundles, referred toas vertical market bundles, from foundationbundles and hardware platforms. Thismakes code reusable and flexible, allowingfor parallelism across platforms.Processor acceleratesperformance, brings it all togetherAs mentioned in previous sections, alow-TDP processor is a key element withineach component of net-centricity, andalso heightens performance. Accordingly,the Intel Atom family offers a highlysuitable solution.PerformanceThe U.S. DoD and its military forcesknow the need for speed can make orbreak a mission – and ensure soldiersurvival. Therefore, a fast processor isintegral. The Atom can meet this needby offering accelerated processing speed:CPU speeds up to 1.6 GHz allow theAtom to support the intensive processingrequirements of modern net-centricmilitary applications.Size and powerEmbedded boards targeted for net-centricwarfare need to be small and low power.Thus, the Atom’s small-form-factorsavvy and low TDP facilitate net-centricmilitary applications. Table 2 showsthe three variations currently available.Each processor has unique capabilitiesand is targeted for specific applications.As the table shows, the Z5xxP stands outwith 2.2 W of TDP. One of this chip’sfeatures that keeps thermal power low isits thermal monitor circuit, which bringsAs demonstrated by the large demand forruggedized, military-grade routers, FastEthernet is also an important way to accessIP. However, traditional data buses willstill play an important role in supportingtraditional military devices and peripheralmachinery. An embedded device will haveto support the latest in Wi-Fi, 1000BASE-TEthernet, or GPS capabilities while stillpreserving high-speed serial ports, USB,PCIe, CAN, and more, making it rich incommunications and I/O.Hardware supported by softwareWith such an extensive list of hardwarerequirements, software is critical to bringthe net-centric system together seamlesslyand effectively. Service OrientedArchitecture (SOA) is a key element. Infact, one of the tenets of the GIG is to useSOA capabilities to increase the flexibilityand portability of applications used innet-centric operations.EnterprisePubliccloudPrivatecloudApplication FrameworkVertical Market Bundles:• Application-specificfunctionality• Compliancy with industrystandardsFoundation Bundles:• Hardware virtualization• Device configuration• System logger• Platform management• Network devicemanagementJava Virtual Machine (JVM)The SOA strategy is data-focused ratherthan hardware-focused. This means moreparallelism as information is readilyavailable to several players with accessto the grid regardless of the underlyinghardware used by each specific player.The SOA approach is very different fromthe traditional approach, which tended tobe more serial, using a sensor or piece ofhardware to communicate serially to adata-gathering station that then dissipatedthe information to another entity.Bootloader / BIOS / Operating SystemEmbedded Hardware Platform802.11 802.15.4 Ethernet Bluetooth RS-485 RS-232 Discrete I/OGPS PC/104 GSM HSDPA CDMA EVDO PCIeFigure 2 | Middleware allows code to be reusable and portable across hardware platforms by separatingfoundation bundles from application-specific bundles.32 May 2010 Military EMBEDDED SYSTEMS

down the temperature when the processorreaches its +85 °C limit. The coolingprocess occurs with minimal effects onapplication performance.To improve efficiency and minimizewaste, Z5xxP also includes a new C6deep sleep mode that boasts a very efficientpower spec of 0.1 W. Table 2 additionallyhighlights the sleep/idle powerconsumption of other processors; thesefeatures are ideal for net-centric militarydevices because they eliminate theneed for bulky cooling components.The end result is an extended lifetimefor the embedded device withoutconcern for unnecessary electrical ormechanical failures.Processor Processing speed Thermal DesignPower (TDP)Table 2 | Comparison of processors’ speed, TDP, and deep sleep/idle power metricsDeep sleep/Idle powerIntel Atom Z5xxP 1.1 GHz – 1.6 GHz 2.2 W 0.5 W – 0.1 WIntel Atom N270 1.6 GHz 2.5 W 0.5 WIntel Atom D510 1.6 GHz (dual core) 16 W 4.5 WHaritha Treadway is a product manager at Eurotech Inc.,responsible for the company’s ARM- and x86-based boardsportfolio. She had 10 years of experience as an engineer in thesemiconductor industry before joining Eurotech. Harithareceived a BS in Electrical Engineering from Cornell Universityand an MBA from Boston College. She can be contacted atharitha.treadway@eurotech.com.Eurotech, Inc. • 301-490-4007 • www.eurotech.comGraphics and visualizationDespite small size and low powerrequirements, military net-centric applicationsstill require high-speed graphicsand visualization capabilities to getmission-critical information to the warfighteron a moment’s notice. The AtomD510 meets this need, with an ultra-fast400 MHz graphics engine integratedinto the CPU.I/O and communicationsIn addition, with extensive I/O support,the military can interface with severaldevices either wirelessly or through standardmeans. The D510 is a good exampleof this, offering speeds up to 1.8 GHzwith dual-core capability. This high-speedI/O support meets the processing crunchrequired by communications, handheld,or radio devices deployed in net-centricwarfare.Software considerationsAtom processors are compatible withan SOA framework and also supporthyper-threading to improve performancein multi threading or multitasking applications,which are often required ofnet-centric military devices.Embedded systems drivenet-centric warfareEmbedded hardware providers can helpthe United States military migrate to netcentricwarfare techniques by makingsure their products meet performanceneeds, provide optimal size and powerconsumption, enable fast graphics andI/O, and provide SOA framework compatibility.Atom-based products – such asEurotech’s Catalyst family of embeddedmodules – meet these needs and are anideal fit for modern net-centric militaryapplications.Military EMBEDDED SYSTEMS May 2010 33

Mil Tech Trends: Battlefield security – from the ground upSecure virtualization combinestraditional desktop OSs and embeddedRTOSs in military embedded systemsBy Robert DayAdvances in software and hardware technologies now make it feasible to use both embedded and desktop operating systemsin a secure military system. Robert examines enablers such as a secure separation kernel and an embedded software hypervisor,then explains uses of desktop OSs in secure military systems.470th Military Intelligence Brigade Public Affairs photo by Gregory RippsAs Intel continues to bring its processortechnologies into the embedded world,an interesting convergence of embeddedapplications with more traditionaldesktop applications is taking place. Formilitary applications, desktop systemsand embedded systems have traditionallybeen separate systems, connected over asecure network (see Figure 1). However,there is now a desire to consolidatemultiple hardware platforms to reduceSize, Weight, and Power (SWaP) whilemaintaining the security that discretesystems traditionally offered.By combining new software and hardwaretechnologies, this consolidation is now areality, without having to sacrifice eitherperformance or security. The softwaretechnology is a secure separation kerneland embedded hypervisor, utilizing theIntel multicore virtualized hardware technology.This software platform becomesa true enabler of modern hardware functionality;however, before examiningthe application of the technology, it isbeneficial to examine the two componentparts of the software.Figure 1 | Traditional systems have physically separate hardware to maintain security and performance.Software component 1:Secure separation kernelA separation kernel is a small, lightweightoperating system that is the lowest-levelconnection to the processor. The separationit provides is not dissimilar to atraditional time- and space-partitionedOS (see Sidebar 1), but it also adds asecure function by enforcing predefinedsecurity policies in areas such as devicemanagement and interpartition communication.Also, the separation kernelitself does not offer traditional OS featuressuch as disk or network access, butit does manage scheduling and memoryfunctions. The advantage of removingmany of the high-level OS features isthat the separation kernel can be keptsmall and efficient, offering real-timeapplication performance and secure,high-speed interpartition communicationusing memory rather than physicalnetworking connections.34 May 2010 Military EMBEDDED SYSTEMS

Safety-critical OSs vs. security separation kernelsFor safety-critical systems, a new class of partitionedRTOSs has been developed in recent years, based aroundan ARINC-653 model. These partitioned RTOSs provide theironclad guarantees of time and space (for example, memory)partitioning, which prevent one software function from interferingwith another. They also enforce strict quotas/budgets ona function’s access to system resources (RAM, CPU time) andrestrictions on which a complement of software functionsmay be active on a given module.As with a safety-partitioned RTOS, a separation kernel’sprimary job is to let developers establish partitions, then toenforce those partitions during runtime. In this sense, aseparation kernel and a partitioned RTOS are very similar.However, there are key differences:• A separation kernel is primarily concerned withpartitioning a system’s data and resources and controllinginformation flow between partitions.• A separation kernel must enforce strict adherence todata isolation, damage limitation, and informationflow policies.• A separation kernel operates in a much more dynamicenvironment than a partitioned RTOS.• A separation kernel must protect against malicious attacks,from without and within.• When combined with hypervisor technology, the separationkernel can run both software applications and other OSs intheir own secure partitions.For high-assurance systems, the Common Criteria evaluationis more complex and hence more costly than a DO-178Bcertification is for safety. Meeting the highest EAL criteriarequires the use of formal methods, where the separationkernel developer proves security properties of the separationkernel. While formal methods have improved significantly fromtheir early days, it quickly becomes intractable to use them onsoftware of even modest size/complexity (a few thousand lines,at most). Consequently, separation kernels intended for themost security-critical applications have fundamentally differentarchitectures from those found in safety-partitioned RTOSs.The functionality of a separation kernel is thereforequite different from that of a safety-critical OS. And theaddition of the hypervisor allows “guest” OSs to sit on topof the separation kernel and give additional functionalityto applications, including Linux/Windows functionality thatcould not be found in a safety-critical OS. Unlike the monolithicsafety-critical OS, the separation kernel and hypervisor canbe separated from the guest OSs when it comes to evaluation,thus reducing the amount of code that has to be evaluatedto the highest levels and helping to implement a system withmultiple levels of security evaluation.Sidebar 1 | There are differences between safety-critical OSs and security separation kernels.In the security world, this smallseparation kernel is the cornerstone ofhigh-assurance systems, offering securitypolicy enforcement and strict partitioning,using a Multiple Independent Levelsof Security (MILS) architecture. Thisallows security engineers to build systemsthat need to be taken to the highest levelof Common Criteria (currently EAL 7)and run applications requiring differentsecurity levels on the same physicalhardware. Many separation kernels arederived from partitioned OSs by removingOS functionality and adding securityfeatures. However, to achieve the highestlevels of evaluation, the software mustalso be proven secure by using formalmethodsanalysis. The separation kernelis the fundamental enabler to the securecoexistence of multiple applications onthe same hardware platform. And, whenunited with an embedded hypervisor, thecombination of desktop and embeddedsystems can be achieved.IT departments to run all their requiredapplications across multiple versions ofserver-based OSs. In the embedded world,the use of hypervisors is not as common.The requirement to run multiple differentversions of an OS on a dedicated embeddedsystem is not as crucial. And therehave been questions over the performanceof running extra layers of software insystems where real-time performance iskey. When a hypervisor and a separationkernel are combined, the ability to bringdesktop and embedded systems togetherbecomes a reality (see Figure 2).Hardware: Desktop OSs in securemilitary systemsWith the use of Intel processors, traditionaldesktop OSs are also being used in manymilitary systems. However, when multiplelevels of security are required, thiscan stop the use of nonsecure desktopOSs. With the introduction of a secureseparation kernel and hypervisor, traditionaldesktop OSs and applications canbe run in their own unclassified partition,thus allowing for the functionality of aknown user interface and applications,without compromising the security of theSoftware component 2:Embedded hypervisorA software hypervisor is a software layerthat allows different guest OSs to resideon a single hardware platform. This technologyis commonly used in the enterpriseor data center realm to allow theFigure 2 | The combination of separation kernel and hypervisor allows multiple OSs to be run securelyon the same physical hardware.Military EMBEDDED SYSTEMS May 2010 35

Mil Tech Trends: Battlefield security – from the ground uprest of the system. Anything that entersinto the desktop partition cannot breachthe secure separation kernel and hencewill be contained in the unclassified partof the system.This software partitioning and virtualizationalso aid in the consolidation of hardwareand the reduction of SWaP, whichis of particular interest in many militaryscenarios. By running separate systemsin their own partitions, and allowing fordifferent OSs and applications to be runin those partitions, there can be a true consolidationof physically separate systemsto a single physical piece of hardware.The use of Intel multicore, virtualizedprocessors allows the merging of aWindows or Linux desktop system witha more traditional Real-Time OperatingSystem (RTOS), and allows the sameperformance and functionality of applicationsas if they were still running on theirown dedicated hardware platforms.An additional feature that is very compellingin regard to this approach is thatof virtual networking. Here, the guestOSs and applications can communicate“virtually” with other guest OSs and applications,even though they are residingin separate partitions. The virtual networklooks to the applications as a real networkport, and so these applications can communicateas if they were two physicallyseparate networked devices, even thoughthe communication is internal. A secureseparation kernel can also enforcesecurity policies to this virtual networkingand dictate which partitions cancommunicate with each other and inwhich direction (see Figure 3).This gives a secure partitioned environmentwith the ability to run multiple guestOSs and applications separated from oneanother on the same hardware. To allownear-native performance while maintainingreal-time determinism and security,hardware virtualization support for bothexecution and memory can be utilizedby the separation kernel and hypervisor.Independent studies performed onthe LynxSecure separation kernel andhypervisor have shown that runningbenchmark applications on a virtualizedLinux OS yields less than a 5 percentperformance degradation as compared tothe same applications running on a nativeFigure 3 | The use of a separation kernel and hypervisor allows desktop OSs and RTOS to reside on thesame hardware platform.implementation of the same Linux on thesame hardware.Another benefit of the migration ofdesktop systems is afforded if theembedded hypervisor uses Intel’sVirtualization Technology. This allowsMicrosoft Windows to be run in fullyvirtualized mode, which requires nochanges to Windows to run on the hypervisor,and a combination of the softwareseparation kernel and the hardware virtualizationgives Windows the impression ithas the whole system, while running inits own secure partition. If no changes arerequired to either Windows or its applications,this speeds the development orporting activity from a stand-alone systemto a secure virtualized system.An example of a MILS solution runningon Intel virtualized hardware isLynxSecure from LynuxWorks. It is asecure separation kernel and embeddedhypervisor that uniquely offers bothpara- and full-virtualization of guest OSs,and maintains real-time performance andMILS security that can be evaluated to thehighest Common Criteria levels. It takesadvantage of multicore Intel componentsto enable high performance even whenrunning multiple guest OSs. MicrosoftWindows can run on the same systemas Linux and RTOS, with each havingits own secure partition and runningapplications at different security classifications.For the next generation of militaryembedded systems, the combinationof LynxSecure and Intel hardware allowsthe ultimate in flexibility of system andapplications while maintaining the highestlevel of security.Robert Day is VicePresident of Marketingfor LynuxWorks,where he is responsiblefor all globalexternal andinternal marketingfunctions. Withmore than 20 years in the embeddedindustry, his most recent position priorto joining LynuxWorks was headingmarketing for Mentor Graphics’embedded software division. Prior tothe marketing role, he held a variety ofmanagement, sales, and engineeringpositions for Mentor Graphics andMicrotec Research, spanning more than18 years in total. Based in San José,California, Robert is a graduate ofThe University of Brighton, England,where he earned a Bachelor of Sciencedegree in Computer Science. He canbe contacted at rday@lnxw.com.LynuxWorks408-979-3900www.lynuxworks.com36 May 2010 Military EMBEDDED SYSTEMS

Mil Tech Trends: Battlefield security – from the ground upDeveloping high-performanceembedded network security applications:A heterogeneous multicore processing approachBy Daniel ProchToday’s network and computing infrastructure is of critical interest to our national security. The vital communicationsystems supporting our multinational forces are valuable resources that share a couple of important characteristics: Theyrequire massive amounts of bandwidth to support our insatiable appetite for IP-based services and they require mechanismsoffering visibility into all data protocol and application layers to ensure network security. Unfortunately, the network appliancescommonly hosting security applications have failed to keep pace with improvements in network performance. However, a newheterogeneous multicore processing architecture can scale to support tomorrow’s throughput needs while providing the abilityto see deeply into network traffic.U.S. Army photo by Staff Sgt. Alex LiceaThe amount of network traffic in today’s wired and wirelessinfrastructure continues to rise at dramatic rates to keep up withvoice/video/data services and real-time military applications. Inboth classified and unclassified military networks, line rates of10 Gbps are commonplace and are expected to quickly grow to100 Gbps in the next several years. These throughputs are largelythe result of new IP-based network-centric warfare applicationslike real-time battlefield monitoring, video surveillance, andfully networked forces. The scalability of the network as a wholeis a major risk to the effectiveness of any network-centric warfareprogram.As network throughputs explode, we also need to be able tointelligently monitor our networks for exploits and to protectconfidential data sources from breaches. An increasing threatto our homeland security is the growing number of high-profilecyber attacks on major military installations, our communicationssystems, government agencies, and financial markets andthe resultant leakage of classified information and personal data.Even Google has been attacked recently in what could be aninstance of state-sponsored corporate espionage. Compoundingthe problem, there are countless other attacks never publicized,but rather hidden by a web of obscurity for obvious confidentialityand national security reasons. The security applications responsiblefor protecting these critical resources need to keep pacewith these increasing network throughputs with even greaternetwork intelligence. Thus, communications equipment mustprovide complete visibility into network traffic at extremely highbandwidths by using content inspection to ascertain the nature oftraffic, not just its destination.Military networks already deploy an array of security applicationsto protect their classified and unclassified resources. These applicationsinclude virus scanning, firewalls, Intrusion Detection andPrevention Systems (IDS/IPS), Distributed Denial of Service(DDoS) mitigation, Data Loss Prevention (DLP), test and measurement,and network forensics solutions. These applicationswork almost entirely by providing Deep Packet Inspection(DPI) and flow analysis, looking for known patterns in networkflows and blocking or recording them. With the need for applicationawareness, security processing, and DPI, the amount ofprocessing power required for these computationally intenseapplications grows exponentially at these increasing line rates.However, these needs for increased visibility, throughput, and38 May 2010 Military EMBEDDED SYSTEMS

network processing power can be met by a heterogenous multicoreprocessing architecture.Heterogeneous multicore processing paradigmsolves paradoxNetwork and security applications can generally be viewed inseveral distinct network architectures. Compute-intensive applicationslike intrusion prevention systems are deployed as activeelements sitting directly on the network wire (inline) processingevery bit of data that traverses the application in real time.These active security appliances need to operate at networkline rates with very low latency. Computation that adds justmicroseconds of delay to network traffic can ruin the effectivenessof real-time end-user applications like sensitive militarytelemetry systems or Voice over IP (VoIP). Specifically,developers typically view 250 microseconds of delay that anyinline network element can add (before end-user applicationperformance degrades) as a high watermark in 1 Gbps networks.Specialized packet, flow, and applicationworkload processingAs shown in Figure 2, a distributed network accelerationarchitecture uses a multi-chip system to achieve maximumperformance and application effectiveness. The three distinctprocessing stages function as shown.Alternately, passive computing elements like network andcomputer forensics systems, intrusion detection systems,honeypots, and vulnerability scanners are not in the directnetwork path, but rather are deployed off a span port, networktap, or mirrored switch interface. These systems are responsiblefor collecting and analyzing terabytes of data from distributedsensors as would be typical in a battlefield node scenario. Thesepassive monitoring devices can offer a thorough understanding ofa network’s topology and which services are available, and scanto assess which vulnerabilities might be exposed on the network.Network appliances deployed in either a passive or active networkarchitecture share a common trait in that they must guarantee100 percent traffic capture across all packet sizes to be effective.Missing any portion of data in a communications streamposes a large threat, making the overriding security applicationineffective.Meeting these performance challenges warrants a new approachto the development of the high-performance systems requiredby the intelligent network. Such systems need to be capableof analyzing traffic at all layers of the OSI model, from thedata link layer (Layer 2) all the way into the applicationspace (Layer 7) while performing this intelligent processingon all traffic at sustained throughputs of 10 Gbps and higher.Achievement of these goals requires specialized and variedprocessing elements, each custom designed for a specific typeof workload computation.A heterogeneous multicore architecture sets a new performancebenchmark for embedded application development thoughseparate and discrete processing elements for packet classification,stateful flow management, and application and control planeprocessing, each with increasingly fine granularity. This architecturetightly couples off-the-shelf Ethernet switch processorsand network flow processor cores with general-purpose multicorex86 systems over a high-speed 40 Gbps, virtualized PCIedatapath. This architecture can be scaled from very low-endsystems up to appliances offering hundreds of Gigabits per secondof packet analysis, stateful flow monitoring, DPI, and applicationthroughput, all with a common software architecture. Accelerateddesigns based on a heterogeneous multicore architecture canenable equipment providers to deliver high-performance, flexiblesystems that are up to four times more efficient than systemsbased on x86 general purpose processors alone with standardNetwork Interface Cards (NICs) as shown in Figure 1.Figure 1 | Designs based on a heterogeneous multicore architectureenable high-performance, flexible systems up to four times more efficientthan traditional x86 systems and standard NICs.Figure 2 | A three-layered heterogeneous processing paradigm uses variedspecialized processors to achieve maximum performance while keepingoverall system costs low.Military EMBEDDED SYSTEMS May 2010 39

Mil Tech Trends: Battlefield security – from the ground up“The mission-critical nature of ourmilitary networks driven by informationcentricwarfare and homeland securitycreates an opposing set of forces.”Ethernet packet processingTo heighten performance levels, off-the-shelf Ethernet switchprocessors are commonplace, offering up to hundreds of Gbpsof configurable packet processing spanning the datalink, IP,and TCP/UDP packet layers. Traffic is classified on ingress andoptionally filtered, cut-through to another network interface, orload-balanced across the Network Flow Processors (NFPs) thatsit logically behind the Ethernet switch processors.NFPs to accelerate higher-layer flow processingNFPs containing a powerful array of microengine RISCprocessors are specialized, multicore devices optimized tooffload burdensome workloads from general-purpose multicoreCPUs. NFPs can handle lower-layer packet processing andaccelerate higher-layer flow and application level processing.This accelerated architecture utilizes the network-optimizedNFP cores for switching and routing, packet classification, statefulflow analysis, DPI, and dynamic flow-based load balancing.Other network processing functions such as TCP termination andSSL offload can also be performed on the NFPs and offloadedfrom the general-purpose CPUs. Traffic can be cleanly structuredfor transmission from the NFP to the general-purpose cores forapplication processing, thereby increasing host performance.Additionally, network flow processors provide hardware accelerationengines for PKI and bulk cryptography to assure line-ratethroughput.PCIe communications path to x86 coresGeneral-purpose multicore x86 CPU(s) in a system are optimizedfor application and control plane processing. From the networkflow processors, packets are passed to the x86 cores across ahigh-performance, virtualization-aware PCIe communicationspath. An efficient zero-copy technique allows the transfers ofpackets directly into user-space application memory bypassingthe operating system kernel, further accelerating application performance.Flows can be pinned directly to specific applicationsor load-balanced across parallel application instances to scaleapplication performance.As shown in Figure 3, through a cooperative set of software APIsbetween the x86 CPUs and NFP cores, the treatment of flowscan also be updated in real-time, offering the ability to changethe treatment of a flow after x86 analysis. This type of flexibilityis essential in situations where a specific portion of a flow is ofinterest for inspection. After inspection is complete, all subsequentpackets belonging to the flow can be filtered or cut through at theNFP layer, which conserves valuable PCIe bandwidth and reducesx86 CPU cycles. Through this heterogeneous architecture, theFigure 3 | Via a set of software APIs between the x86 CPUs and NFPcores, the treatment of flows can be updated in real-time. This flexibility isessential when a specific portion of a flow is of interest for inspection.general-purpose multicore processors can focus on the computeworkloads they are best suited for such as behavioral heuristics,Perl Compatible Regular Expression (PCRE) processing, contentinspection, and analysis or other similar applications.Intelligent Networks at 40 GbpsThe mission-critical nature of our military networks driven byinformation-centric warfare and homeland security creates anopposing set of forces. Networks need to continue to scale to meetexponentially growing bandwidth demands, and enterprises needthe ability to effectively monitor these networks at all packet andcontent layers with stateful network intelligence. To meet theseneeds, a distributed, multi-chip, heterogeneous multicore architectureis required, providing specialized workload processing toeffectively scale applications to 40 Gbps and beyond.Daniel Proch is director of productmanagement at Netronome responsiblefor their line of network flow engineacceleration cards and flow managementsoftware. He has 14 years of experiencein networking and telecommunicationsspanning product management, CTO’s office,strategic planning, engineering, and technicalsupport. Previously, Daniel was with FORE Systems andremained with the organization through acquisitions by Marconiand Ericsson. Daniel has a BS in Mechanical Engineeringfrom Carnegie Mellon and an MS in Information Science andTelecommunications from the University of Pittsburgh. He canbe contacted at daniel.proch@netronome.com.Netronome724-778-3290www.netronome.com40 May 2010 Military EMBEDDED SYSTEMS

Editor’s Choice ProductsEditor’s note: Military Embedded Systems is “hip” to the whole Web 2.0 social networking revolution. While wedon’t know which of today’s buzzy trends will last, we’re going to start including links to vendors’ social networks, whenprovided. You can also reach us on Twitter, Facebook, and LinkedIn ... and that’s just for this week. Next week there’llundoubtedly be more new sites.Adapters preserve the legacy display investmentSometimes the introduction and then proliferation of new technology wares yields angst for the budget-conscious consumer orDoD department heads looking to stretch defense dollars as far as they can. While Mini DisplayPort and DisplayPort rapidly become thede facto display-to-computer interfaces in new mobile laptop or desktop computers, many already-owned-and-paid-for displays and monitorsonly support older video connections including VGA, HDMI, or DVI. However, StarTech.com’s Mini DisplayPort adapters link old and new –and eliminate the need to purchase a new monitor or display. Sounds perfect for a military command and control center looking to next year’sbudget and tightening the monetary belt in the meantime.And the investment-preserving adapters provide many notables. Among them: Mini DisplayPort adapters offer display resolutions up to 1920 x 1200, and are software-free for easyimplementation. In addition, HDTV support renders resolutions up to 1080p. And the plug-and-play Mac- and PC-compatible adapters are available in three variants: Mini DisplayPort toDVI, Mini DisplayPort to HDMI, and Mini DisplayPort to VGA.StarTech.com • RSC# 45104 • www.startech.comStorage systemsatisfies missioncriticalneedsStorage is everywhere, and with good reason.The world would be chaotic without ways to storedata or even personal collectibles, and work wouldhave to be repeated incessantly to replace fleeingdata or belongings. On the mission-critical side,Themis Computer’s RES-XStore storage systemcan help satisfy the thirst for storage by providingharsh-environment storage capacity of 12 TB.Tucked into a 1U (17" or 432 mm) chassis,RES-XStore utilizes four hot-pluggable canisterswith a total capacity for twelve 2.5" SATA or SASdrives. A RAID controller supporting either JBODor single disk at levels ranging from 0 to 60is also provided, as is RAID volume managementand RAID out-of-band or in-band management.(Out-of-band management is additionally affordedvia Ethernet.)Storage system to host server communicationis facilitated by PCI Express x8 and an add-inhost adapter. And since this communication isparamount in mission-critical scenarios, RES-XStorecan withstand operating shock at 25 g, 20 msand vibration at 3.0 Grms, 8 Hz to 2,000 Hz.It also complies with MIL-STD-810G, and, sportinga SATA HDD, RES-XStore’s operating temp is0 °C to 60 °C. And another plus: Its modulardesign simplifies service and upgrades.Themis ComputerRSC# 45086www.themis.com42 May 2010 Military EMBEDDED SYSTEMS

FMC clock generator boosts high-frequency data samplingA high-performance, reliable clock generator is an integral part of high-frequency data sampling applications, andCurtiss-Wright Controls Embedded Computing’s FMC-XCLK2 quad-channel clock generator FMC card provides that … and more.Designed for sampling apps such as radar, SIGINT, spectral analysis, SDR, and Electronic Counter Measures (ECM), the FMC-XCLK2also, importantly, eases FPGAs’ integration into embedded systems. The simplifying factor is its FPGA Mezzanine Card or FMC(VITA 57) form factor, designed to ease FPGA I/O at half a PMC’s size.The FMC-XCLK2 supports 50 MHz to >2 GHz RF output frequencies and provides a source for synchronizing and clockingI/O. In fact, the FMC card offers up to 4x phase matched outputs (two differential) – ideal for synching multiple I/Os – alongwith 10 MHz master reference output, ultra-low jitter, and selectable external/internal 10 MHz reference including RF outputfrequency multipliers. Input levels for the external 10 MHz reference are 50 Ohm AC coupled, with -5 to +5 dBm recommended.Meanwhile, the external RF clock input supports the same range. Effective in high-performance I/O wares like ADCs and DACs, FMC-XCLK2 is manufactured in either ruggedconduction-cooled or air-cooled variants.Curtiss-Wright Controls Embedded Computing • RSC# 45097 • www.cwcembedded.comPCI2PMC adapter joins two worldsIt almost looks like a texting acronym, but it’s not one a teen would use. Rather, “PCI2PMC” is the model numberof Dynamic Engineering’s PCI-to-PMC adapter, enabling PMC card installation into either a half-length or standard PCIslot. The adapter actually measures 6.600" per the PCI specification, meaning it’s an adaptable half-length ware withuniversal voltage. Fully loaded, PCI2PMC provides a passive design.With 33/66 MHz bus operation, the adapter renders 32-/64-bit data transfers. PCI2PMC also supports 3.3 and 5 VPCI bus signaling, and +3.3, +5, +12, and -12 V can be supplied from the PCI backplane to the PMC with an optionaljumper for PCI or a regulator at 3.3 V. Convenience in connection is provided via front- and rear-accessible connectorsincluding a rear DIN64/SCSI connector. Compatible with third-party or Dynamic Engineering PMCs, the adapter’scut-out design enables increased airflow, and the unit can operate in the industrial temp range of -40 °C to + 85 °C.Meanwhile, build options for the 4.6 million-hour MTBF adapter include ditching the standard DIN for a SCSI connector, adding PrPMC clocking, conformal coating, and JTAG for convenientdebugging, or an RoHS version. Software is low-maintenance, as the passive adapter merely enables usage of the original PMC software.Dynamic Engineering • RSC# 45099 • www.dyneng.comBus analyzer GUI’s got the lookFace it. Looks do matter: not just on the runway or new car lot, but certainly also when it comes to pitting character-based busanalyzers against modern Graphical User Interfaces (GUIs). Accordingly, one of the knockout features on GE Intelligent Platforms’version 7 of its BusTools-1553 bus analyzer software is its new “intuitive” GUI. The GUI simulates, tests, and analyzes 1553 data bustraffic on CompactPCI, PCI, VME, VXI, PC/104, and PC/104-Plus form factors, among others. The GUI interface hastens bus trafficanalysis and monitoring and facilitates fast message modification/creation. It also enables simultaneous multiple-bus control in additionto error detection/injection and speedy filtering for either recorded or live displayed data.And BusTools-1553 version 7 has even more new tricks up its sleeve. One is the Dynamic Bus Monitor stop/start feature, enablingusers to achieve efficient on-the-fly 1553 bus traffic routing. Another is a one-page bus list editor, replacing the multi-page editor andpermitting users to view every bus traffic message within a single window, then organize them quickly onto a highly readable one-pagelist. And finally, BusTools-1553 version 7’s Selective Data Watch feature lets readers choose different data words from any busmessage to identify elusive system issues, thanks to integrated high/low limit checking, automatic limit event logging and corresponding snapshot feature, and DDE output.GE Intelligent Platforms • RSC# 45098 • www.ge-ip.comALM system offers one-stop-shop for safety-critical appsLife-cycle management is paramount in the defense industry, and not just at the component level. The application level,particularly for safety-critical apps like avionics, also holds its challenges in dovetailing all project phases. Enter Embed-X, an“end-to-end” Application Lifecycle Management (ALM) system from LDRA and Visure Solutions. Embed-X integrates the entiresoftware engineering process, from project management, to requirements management, to architecture, code creation, softwareconfiguration, and finally testing. Now requirements, for example those for DO-178B, can be found and used dynamically withinthe system alongside the tangibles from all other project phases, rather than being isolated and located elsewhere.To maintain full phase integration, Embed-X reports on the progression of development and enforces inter-phase developmentprocesses. LDRA and Visure Solutions estimate that defense and avionics vendors often experience 2x budget overruns, but suggestthat Embed-X’s one-stop-shop efficiency could yield as much as 50 percent in cost savings. Where does its efficiency come from?Safety-critical requirements tracing is executed via dynamic and static analysis all the way through to testing and verification. Additionally, Embed-X is compliant with MISRA, securitystandards including the Homeland Security Agency’s Common Weakness Enumerations (CWE) and Cert C, plus safety-critical standards including DO-178B.LDRA • RSC# 45100 • www.ldra.com / Visure Solutions • RSC# 45101 • www.visuresolutions.comMilitary EMBEDDED SYSTEMS May 2010 43

Editor’s Choice ProductsFPGA starter kit provides the whole packageFPGAs are prevalent in all sorts of military embedded applications and systems these days, and form factors such asAdvancedTCA, MicroTCA (collectively referred to as “xTCA”), and AdvancedMC (AMC) are increasingly following suit. Thing is, howare FPGAs integrated with these historically telecom form factors? Pigeon Point Systems’ Module Management Controller (MMC)Board Management Reference (BMR) Starter Kit provides the key by rendering all firmware needed for FPGA development, withthe aim of quicker deployment of AMC products. The kit also includes a benchtop management controller development board inAMC form factor style and a full SmartFusion FPGA design tucked inside a Libero Integrated Design Environment (IDE). A productionlicense and “comprehensive documentation” round out the entire picture. And xTCA developers will be happy: Pigeon Point makestwo benchtop variants of the kit: One for MicroTCA, one for AdvancedTCA.Meanwhile, the kit’s foundational product is the SmartFusion intelligent mixed-signal FPGA product including built-in flash.What makes SmartFusion unique is its composition: an FPGA, programmable analog, plus a 32-bit ARM Cortex-M3-based Microcontroller Subsystem (MSS) at 40 MHz. With theARM processor experiencing zero load while SmartFusion renders advanced analog processing, the processor is left to execute xTCA analog sensor monitoring. Now that’s what wecall efficient.Pigeon Point Systems • RSC# 45102 • www.pigeonpoint.comFind errors your way –ahead of timeIn safety-critical systems, failure is simply not anoption. Such thinking was surely the impetus behindaicas GmbH developing their Java-based VeriFluxautomatic static code analysis tool. (The tooladditionally works well on non-safety-critical Javacode.) While testing always mandates executing yetanother run for each code path, VeriFlux can analyzeall code paths simultaneously. Not only that, VeriFluxworks to prevent runtime exceptions by pinpointingeverything from null pointer errors to arithmeticissues. Deadlocks, especially in multicore programs,can be flushed out with the static analysis tool, andrace conditions can be prevented as well. And if thesystem carries dead weight (aka “dead code”),it won’t get past the tool’s highlighting.One thing that makes VeriFlux particularly intriguing,however, is its configurability: It takes orders fromthe user, meaning that it can be “told” to focus oncertain aspects of the analyzed code and to providea briefer summary for the remaining code. It can alsobe used anytime in the development cycle, includinganalyzing incomplete programs or just a singlemethod. Meanwhile, other notables include the tool’sRTSJ memory that validates all scoped memoryassignments, in addition to VeriFlux’s automatic simplereflection or user-programmed complex reflection.Library modules are also included in the tool toprovide ease of use.aicas GmbHRSC# 44362 • www.aicas.com44 May 2010 Military EMBEDDED SYSTEMS

GbE switch card aids soldiers and their networksOut in the harsh environmental terrain of Afghanistan, situational awareness is paramount to mission success and survival.One key enabler is network-centricity, allowing soldiers to receive strategic or protective orders from commanders and other personnelin the timeliest and most straightforward way. We’re going to guess that developers at Parvus had just such a scenario on their mindswhen they developed the new COM-1268 rugged PC/104-Plus GbE switch card. Designed to heighten IP network-centricity andsituational awareness, the 10-port board provides robust network performance even under the most demanding thermal andshock/vibe conditions.Supporting both IPv4 and the later-and-greater IPv6 Quality of Service (QoS) traffic prioritization, the Layer 2 GbE switchsurvives – and thrives – in a wide temperature range of -40 °C to +85 °C. It also supports Simple Network Management Protocol(SNMP), Virtual Local-Area Network (VLAN) trunking, and Rapid Spanning Tree (RSP) redundancy. And it is vended as a stand-aloneboard … or not: It can also be integrated into a mission computer platform, mobile router system, or ruggedized COTS switch subsystem, for example. And there’s even more good news:The COM-1268 does not need an additional processor for board operation and can also be used in non-PC/104 systems.Parvus Corporation • www.parvus.com • RSC# 44031Show me the data ... in SWaP-savvy styleData gathering, formatting, and transmission are imperative to fostering mission success and soldier safety, particularlyfor manned military aircraft, for example. And the Aitech NightHawk RCU [Rugged Controller Unit and Data ConcentratorUnit (DCU)] is a technology designed to that end – and more. Based on the compact Intel Atom N270 processor runningat 1.6 GHz with 2 GB DDR2 SDRAM, the ultra-rugged 4.5-lb unit is designed for SWaP-conscious airborne or groundvehicles in the defense and industrial arenas. With a slim profile sporting natural convection/radiation cooling, thecompany reports that NightHawk RCU “dissipates over 22 W at +55 °C in free air, or at up to +71 °C with an optionallow-pressure fan.” A triad of rugged MIL-DTL-38999 connectors additionally boosts robustness.Specifically suited for harsh-environment chemical, climatic, mechanical, and electrical apps, NightHawk RCU’s datastorage is accomplished via a standard offering of 4 to 8 GB SSD or an option to increase to 250 GB SSD. Accoutrements include standard I/O interfaces: dual GbE, stereo audioin/out, RS-232/422/485, PS/2 ports, video graphics, and many more. Meanwhile, NightHawk RCU’s options include WAN wireless radio or Wi-Fi, to enable the RCU to execute periodicdata monitoring or logging to a home base. A second option is a preformed cable set, which makes the unit prototype-ready by merely plugging it into Ethernet and a standard keyboard,mouse, and monitor.Aitech Defense Systems, Inc. • RSC# 44263 • www.rugged.comWIN-T routerhits the streetsMany technologies aredeveloped for a particular militaryprogram, but then what? Sellit as-is to defense and civilianmarkets? Yes ... if you’reJuniper Networks, which recentlyreleased its LN1000 mobile securerouter – used in the U.S. Army’sWarfighter Information Network-Tactical (WIN-T) program – for public consumption. (Well, at least for industriessuch as defense, public safety, utilities/energy, smart grid, and others.)Available in the VPX form factor (4" x 6" x .85"), the 1.5-lb LN1000 providestransmission of data, video, and audio traffic for data aggregation, surveillance,or comms apps. But the notable here is the mobile security, which sounds like anoxymoron but isn’t. Indeed, LN1000 securely interconnects platforms such as remotemonitoring or sensor stations, UAVs, and so forth to their operations centers or centralcommand. Also afforded are resiliency in networking and high performance with lowpower consumption: a mere 35 W. Designed for tough environments, this mobilesecure router withstands temps from -40 °C to +85 °C. And deployment onto existingplatforms is no problem either: LN1000’s conduction-cooled design eliminates theneed for external power.Juniper Networks, Inc.RSC# 45096www.juniper.netEditor’s Choice Products are drawn from OSM’s product database and press releases. Vendors may addtheir new products to our website at http://submit.opensystemsmedia.com and submit press releases athttp://submit.opensystemsmedia.com. OSM reserves the right to publish products based on editors’discretion alone and does not guarantee publication of any product entries.Military EMBEDDED SYSTEMS May 2010 45

Crosshairs EditorialEmbedded SystemsConference portendsCOTS’ futureBy Chris A. Ciufo, EditorEvery year CMP’s Embedded Systems Conference offers aglimpse into technologies likely to find their way onto tomorrow’sbattlefield. At the close of this week’s ESC (ending 29 April 2010),there were four key markets that countless vendors emphasized:Android, industrial, medical, and military. Yeah, that last one surprisedme, too. Rarely do software, IC, and network infrastructurevendors talk about the military market. But unlikely companiesfrom Mentor Graphics to Dell talked about their defense marketplans. After meeting with nearly 50 vendors, here are some COTStrends to watch out for during the next 18 to 24 months.Smartphones; videoWarfighters want the same rich media and Internet connectivityon the battlefield as they have in their off time. Hence, UASvideo feeds, Blue Force Tracking, and moving GPS maps makesense on portable handhelds. And DoD brass wants to give it tothem. I discovered at least two contracts – one from the NSAand one from DARPA – aiming to create Android-based andmilitary-secure iPhone-like handsets. Already, companies likeThales offer secure BlackBerries with https and IPsec security.And consumer devices like Apple’s 3GS iPhone make the videoexperience so compelling that it’s likely all dismounted soldierswill eventually tap into some data stream, somewhere.Cloud computing and “point-of-the spear” processingNot all battlefield computers need heavy horsepower. Thin clientdevices relying on ARM Cortex M3 or QorIQ CPUs connectedto fat pipes and rack-mount servers such as AdvancedTCA or2U rigs from Emerson, RadiSys, or Z Microsystems would becheaper, lighter, and use less power; they would also still provideaccess to information from elsewhere in the GIG. This is thecloud computing model first theorized in X-Windows a decadeago. Intel’s MIDs vision maps perfectly to many battlefield sensorsand embedded platforms to realize cloud computing. Thenagain, some devices such as UAS payload sensors or commsjammers need all the performance they can get in the local box.New DSP offerings in Altera Stratix V devices or Xilinx Virtex-6FPGAs bolted to Core i5 Intel CPUs with Quick Assist bringmassive IOPS to low-power embedded platforms.Fatter pipesThe digitized battlefield needs node connectivity, whether viasatellite, mesh networks such as SNAP from Synapse, or an aloftUAS relay station. But routing video everywhere and to everyonerequires more bandwidth. Trouble is, current technologies aremaxed out, and forklift upgrades are cost-prohibitive. Instead,COTS technologies like JPEG2000 compression, H.264, anddeep packet inspection will reduce the amount of data beingpiped around the battlefield. New CPUs from ARM, Freescale,Intel, and VIA are all optimized for one or more of these datareduction schemes.Application Life-cycle Management (ALM)Not a new technology, ALM is moving into defense system designsthrough offerings such as LDRA’s TBReq and Mentor Graphics’new ReqTracer. Tools like these collect and track requirementsmanagement snippets, making sure that complex, multiyearprograms meet evolving specs. Previously, myriad SOW CDRLswere (at best) tracked via complex, iterative Excel spreadsheets.SecurityWe mention them often in Military Embedded Systems, yethigh-assurance hardware and software are still rare in all butthe most classified of comm systems. But with every embeddedsystem now a node on someone’s network, information assuranceto protect assets from cyber attack is becoming a baselinerequirement. Secure operating systems meeting (or soon to bemeeting) Common Criteria, MLS, or MILS are available fromGreen Hills, Wind River Systems, and LynuxWorks. For dataon-the-fly, Mocana provides security assurance.Safety-critical and MILSThese two terms are often used together but are not synonymous.Safety-critical specs such as DO-178B (soon, C) and DO-254provide certification that code and hardware behave predictably.The same RTOS vendors listed above offer safety-critical offerings,as does AdaCore with their Praxis partnership. MILS, on theother hand, will soon extend beyond just security. How? RTOSvendors have revamped their hypervisors to provide protectedpartitions, which now offer the added benefit of running a guest OSin a protected partition. Vendors at ESC including Enea demonstratedhow even a retail Windows XP installation could run in itsown partition and if crashed or compromised, wouldn’t affect therest of an embedded system. Partitioned environments also allowlegacy operating systems to be moved forward into modern, multicorehardware. Lastly, LynuxWorks demonstrated secure Windowsmachines where the Windows OS is run atop a MILS hypervisor.Small-form-factor shoeboxesI’ve talked about this for years. It’s where a vendor such asWinSystems, ADLINK, Parvus, or others build a custom smallform-factorbox with the “right” amount of I/O and processingpower for the system. Despite the best efforts of consortia likeSFF-SIG, PICMG, or the PC/104 Consortium, DoD designerscare less about the card inside the box than about the box itself.And since shoeboxes are inexpensive and available in myriadruggedization levels, they can often be tossed out during the nexttech refresh.And lastly, more with lessRight now, the industry is girding for reduced DoD budgets asO&M sucks money from RDT&E, despite President Obama’s$708B request (Feb10) versus $690B in FY10. Worse, asSECDEF Gates puts the brakes on underperforming programs,primes expect to have to rejustify many line items. That meansfunding delays that ripple down to the COTS industry. The goodnews is: This is deja vu all over again. COTS has always been theway to get the best tech to the front, fast.Chris A. Ciufo, cciufo@opensystemsmedia.com46 May 2010 Military EMBEDDED SYSTEMS

GEIntelligent PlatformsWelcome to theKnowledge Bank.A wealth of free COTS white papers and tutorials.Information may be the ultimate competitive advantage, and that’s why we offer youfree access to our extensive COTS white paper library. With dozens of papers currentlyavailable for download, you’ll find information ranging from ARINC 429 and MIL-STD-1553tutorials, to an examination of the latest Intel ® Core i7 processors. Who knows what greatideas you may find in one of our papers? It could be the idea that gives you an importantcompetitive edge in developing and delivering the solutions your customers are demanding.www.ge-ip.com/cots-papers© 2010 GE Intelligent Platforms, Inc. All rights reserved.All other brands or names are property of their respective holders.

SPEED YOUR TIME-TO-MARKETYOUR TOTAL SOLUTIONS PARTNER1SignalAcquisitionAnalog & DigitalI/OADC510 (FMC) &FMC-XCLK22 Radar Processing3Digital Signal ProcessorsCHAMP-AV5Flight ControlMulti-Platform MissionComputersMPMC-9350 &MPMC-9310 (VPX,VME or CompactPCI)Do you need innovative solutions or complete integration services for high-density data processing now? Fromboard-level products to fully integrated systems, we provide advanced technology solutions that perform underthe most rugged operating conditions. Speed your time-to-market and lower your overall program developmentcosts utilizing our leading edge, commercial-off-the-shelf products, or modified COTS (MCOTS) and engineeringservices. Ask your representative about the new VPX System specification, OpenVPX.4GraphicsDisplay5Video Input & OutputXMC-710SystemConnectivityEthernet SwitchesVPX6-684 FireBlade6Command &ControlSingle BoardComputersVPX6-187 &VME-1905sales@cwcembedded.comcwcembedded.comBOARDS & SYSTEMS