BWI - ETH ZÃ¼rich - Information Management - ETH ZÃ¼rich

BWI - ETH Zürich 

Trends and Directions in IT and 

Information Management 

Innovative Technologien für „Event Driven“ IT Infrastrukturen 

Francis Kuhlen 

Managing Director IBM 

Information Management | Zürich, 18. Juni 2007

BWI – ETH Zürich 

Accelerating Advances in Technology 

1E+12 

$1000 Buys: Computations per second 

1E+9 

1E+6 

1E+3 

1E+0 

1E-3 

1E-5 

1900 1920 1940 1960 1980 2000 2020 

Source: Kurzweil 1999 – Moravec 1998 

2 

Information Management | Zürich, 18. Juni 2007 

© 2007 Francis Kuhlen, IBM


Accelerating Advances in Technology 

1E+12 

$1000 Buys: Computations per second 

1E+9 

1E+6 

1E+3 

1E+0 

1E-3 

$, €, CHF 

1E-5 

1900 1920 1940 1960 1980 2000 2020 

Source: Kurzweil 1999 – Moravec 1998 

3 




The Internet and World Wide Web 

1500 

$6000 

Users/Buyers/Devices Millions 

1000 

500 

Users 

Buyers 

Devices 

eCommerce 

Spending 

eCommerce Spending $Billions 

0 

2000 2001 2002 2003 2004 2005 2006 

$0 

4 




The Event-Driven World 

IPv6 Map 

VoIP, VoD, IMS, etc. 

RFID-enabled Checkout 

Intelligent Oil Field 

ENRON Chairman 

indicted…….. 

Surveillance 

RFID-enabled Scanning 

• Billing 

• Security 

• Fraud alerts 

• Retrospective 

processing 

5 

Fraud/Compliance 


Location Based Service (Traffic) 



The Event-Driven World 

The accelerating need to handle large volumes of time-dependent 

events will give rise to new classes of middleware, programming 

models, and tools 

● There is a growing need to monitor, capture, process, and store massive volumes of 

time-dependent events 

• Smart sensors, RFID, program trading, fraud management, risk and compliance, intelligent oil field, location 

based service, logistics, presence (SIP), in-line analytics, etc. 

● An increasing number of companies are addressing the needs of the time-dependent 

infrastructure market 

• Developing event engines that route, transform and derive events from multiple streams 

• Delivering content management solutions supporting large volumes of time-dependent data 

• Extracting event information from applications (ERP, CRM, etc.) and sensors 

● Standards are emerging 

• Real-time Specification for Java (JSR 1), Distributed Real-Time Specification (JSR 50) 

• Web Services Notification and Web Services Eventing 

• OMG Data Distribution Service (DDS) and IEEE 1516 for battlefield simulation 

● Middleware will evolve to deal with the throughput and time-dependent needs of the 

event-driven world. New programming models and tools will also emerge 

• Data integration will evolve to event integration 

6 




Any Network 

Provide optimized access seamlessly to user or machine over any network 

Personal Space 

(Office, Briefcase, 

Person, Broadband) 

On-Campus / 

Public (Office, 

SOHO, Airport, 

Hotel, Coffee Shop, 

Broadband) 

City, Community 

(Last Mile, Remote 

Coverage, Fixed and 

Mobile Broadband) 

Cellular/ PCS /Satellites 

(Miles / Regions, National, 

Continents, Fixed and 

Mobile Narrowband) 

PAN 

LAN MAN WAN 

7 




Future Networks 

Data, voice and multimedia will be carried over a heterogeneous 

physical network running IP 

● Supports large number of 

devices… 

• Wireless communicators: Cell phones, 

PDA’s, pagers, etc. 

• Interactive “smart” sensors (health monitors, 

environmental sensors, etc.) 

• RFID tags 

● “True” mobile computing will be 

enabled 

• Complete range of service (internet, TV, 

VoIP, etc.) 

• Self-configuring 

• Seamless roaming 

• On demand remote storage 

Location-based 

Services 

PSTN 

Gateway 

Cellular 

Radio 

PAN 

Base station 

Edge of Network 

Services 

Server 

Enhanced IP 

Core Network 

PAN 

WLAN 

Smart 

Sensors / 

RFID tags 

On 

Demand 

Storage 

Distributed 

storage 

Intelligent network 

elements 

Access 

Router 

Access-point 

SAN 

8 




Event Driven: Pervasive/Ubiquitous Computing 

● Smart Interconnected Devices 

● Multimodal User Interfaces 

● Sensor-based Applications 

9 




Personal Mobile Hub 

Open platform for 

• Providing open interfaces to service 

infrastructure to appliances, accessories, 

sensors, etc. 

• Extending (third party) services & 

applications to the mobile user 

• Enabling (third party) services & applications 

provisioning on the mobile device 

• Running on Standard OS 

And offering the following advantages 

• Short time-to-market for new applications 

based 

on specialized, tailored devices 

• New functionality inexpensive to attach 

• Opens up niche market 

Applications 

• Entertainment, education, logistics, 

maintenance, monitoring, healthcare, 

well-being, …. 

• Bluetooth 

• 802.11 a,b,g 

• UWB 

Near 

Far 

• GSM/GPRS 

• UMTS 

• 802.11 a,b,g 

• UWB 

Local devices 

(sensors, appliances) 

Personal 

Mobile 

Hub 

Network Services 

10 




Personal Mobile Health Hub 

Peekflow 

Body Weight 

Activity 

Inhalation 

Registration 

Heart Rate 

Injection 

Registration 

Blood 

Pressure 

Glucose 

Auscultation 

ECG Monitor 

Pill Dispenser 

11 




Network Architecture Proposal for Cars 

Broadcast 

Digital 

Car Radio 

Mobile Computer Platform 

Software 

Application 

Speech 

Processing 

Wireless 

Mobile Phone, 

Pager 

Positioning 

GPS 

Flash 

RAM 

Embedded 

Power PC 

Operating System 

Java Virtual Machine 

Navigation 

Internet 

E-mail 

Smart Card 

eCommerce 

Other 

Smart Card 

Reader 

Gateway & 

Communication Interfaces 

Automotive Bus 

12 








13 




Scenarios and Applications 

Equipment Equipment 

Sensors Sensors and and 

Controls Controls 

Industrial 

Automation 

Automated Automated 

Data Data 

Collection Collection 

Chemical 

Industry 

Remote 

Metering 

Location 

Positioning 

Distribution & 

Retail Industry 

Remote Remote 

Diagnostics 

Diagnostics 

and and Service Service 


Automotive 

Private 

Security 

RF ID Tags 

Security Security 



Electronic 

Health 

Context 

Awareness 

Asset 

Management 

Ambient 

Intelligence 

Lighting Lighting 

Monitors Monitors and and 


Smart 

Home 

HVAC HVAC 



14 




RFID Tag Characteristics 

● Frequencies 

• ~100kHz access control, animal ID 

• ~10 MHz smart cards 

• ~1000 MHz toll roads & item management 

• ~2.45 GHz item management 

● Power 

• Passive, semi-passive (battery assisted), active (battery powered) 

● Capabilities 

• Read-only, write-once (identity tags) 

• Read-write (multiple), user memory 

• Act as reader 

• Encryption, sensors, ... 

‣Above factors influence price, read range, life time, and kind of data 

cllected/stored 

15 




EPC Tag Classes 

16 




A Simple Sensor Network 

Challenge: end-to-end system solutions 

SU 

SU 

SU 

data aggregation 

& abstraction 

publish / subscribe 

SU 

SU 

SU 

Sytem Components: 

•Middleware 

•Servers 

•Custom designed units 

SU 

AU 

BS 

Server 

Server 

BS 

SU 

SU 

SU 

SU 

Systems aspects: 

•Sensor diagnostics 

•System diagnostics 

•Remote overview 

•Remote configuration 

•Remote SW updates 

•Sensor data access for 

business applications 

SU 

BS 

further data aggregation 

business applications 

BS 

AU 

actuator unit 

SU 

SU 

SU 

severe power consumption 

constraints if wireless 

SU 

AU 

SU 

BS 

sensor unit 

base station 

17 




The Challenge . . . 

Industrial 

Sensors 

Internet 

Scalability problem 

due to I/O limitations 

Environmental 

Sensors 

Data, Events, Messages … 

Application Server 

Telematics Applications 

Data or Events Generated 

By Telco Equipments 

18 




Directional Shift in Network Traffic 

The massive deployment of smart, networked sensors will dramatically 

affect network volume and traffic patterns 

Client 

Server 

Client 

Server 

Client 

Server 

Sensors 

Server 

1:50 

1:1 

1:1 

100:1 

Web Server 

VoIP 

Conversation 

P2P File 

Sharing 

Sensor/RFID 

System 

● Traditionally, client requests are accommodated by caching 

● In future, computation will move to the edge of the network to aggregate, 

synthesize and filter data 

19 




The Solution: Edge Server Software 

Server 

Delegating parts of the 

application to the edge 

Application 

“edgification” 

Clients 

Booster Box 

Application Code 

Modules 

Infrastructure to 

1. deploy part of the code at 

the edge of the network 

2. relay processed information 

to/from the server 

20 








21 




A Short History of User Interfaces on Desktop 

From green screen to web portals 

Windowed Desktop 

Visual 

composition of 

several 

applications on a 

screen 

Portals 

Visual 

composition of 

multiple web 

sites into one 

view 

• Evolving Web Services 

Interfaces 

• Integration of multiple 

information sources and 

functions 

"Green Screen" 

Single 

screen per 

machine 

Single 

application 

at a time 

Browser and Web Site 

Integration of local & 

remote resources 

22 




User Interaction Becomes Multimodal 

Smart Environments 

Voice & Pointing based interaction 

from a distance (e.g. JestureTek) 

Microphone/Camera arrays 

Person tracking 

PC/Desktop 

Head / hand as a mouse (with 

well-controlled background, glove) 

Gaze tracking 

Conversational interaction 

Ink character and word input 

3D articulated body gestures 

Multi-scale analysis of user position, orientation, 

facial features, gestures 

Steerable user-centric interfaces 

Multimodal interaction, person id 

Handwriting capture/reco without touchsceen 

3D gesture based manipulation without glove 

Monitoring of user focus of attention and user 

reaction 

3D head model capture for facial interface and 

virtual conferencing 

Natural, multimodal (voice, GUI, ink, visual) 

Mobile device 

Laser projected keyboards 

(e.g. VKB/Siemens) 

Video capture of face 

Conversational interaction 

Ink character input, messaging 

Increasingly seamless transition between 

PDA/mobile, PC, smart space 

3D head coding for transmission 

audio-visual speech recognition 

Voice+pen+gesture interaction 

23 




InfoScope 

● Combines mobile computing with augmented reality 

● Information augmentation system based on recognition of 

objects in scene images 

● Provide relevant information to travelers, mobile 

professionals, wherever they are and when they need it 

• Automatic text translation 

• About buildings 

• About products 

• About events 

• … 

24 




InfoScope – Automatic Sign/Text Translation 

Client GSM/GPRS Server 

Scene Text Detection 

Yongye Shaola Meat Supermarket 

Character 

Segmentation & 

Recognition 

Provide Fresh Meat Every Day 

Language 

Translation 

Text Augmentation 

25 




InfoScope – Information Augmentation in the City 

Client 

GSM/GPRS 

Server 

Object Detection 

Building Recognition 

Information Retrieval 

Information 

Augmentation 

26 




Page 6 

Examples of Time-Dependent Events 

Internet packet data – Handle huge volumes of intelligent messaging, voice, and video 

over IP networks 

RFID – Track location, count, and flow of parts in global supply chains 

Plant floor automation – Monitoring people and equipment in buildings and plants 

Real-time fraud detection – Monitor transaction authorization systems to identify 

fraudulent behavior and generate real-time alerts before the transaction is approved 

Surveillance – Monitor people in airports for suspicious activity and generate timely alerts 

Intelligent oil field – Generate early warnings to initiate oil well shutdowns by using 

signals and complex analysis from multiple sensors 

Financial Services – Exploit time-dependent opportunities in currency trading and index 

arbitrage 

Health – Detect early warnings of large-scale disease pandemics 

Space – Observe corona mass ejections to obtain early warning of disruptions in global 

weather patterns 

Autonomic – Autonomic monitoring and management of large-scale, virtualized systems 

27 




RFID-Enabled Supply Chain 

RFID enables automated monitoring, routing, and inventory management 

of transportation vehicles, pallets, cases, items, etc., across supply chains 

Manufacturer 

Distribution 

Centers 

2005 WW Statistics: 

500 million pallets per year 

100 billion cases per year 

2 trillion items per year 

Retail Store 

28 




Surveillance 

Example 

Stamford, CT 

Newark Airport, NJ 

Sensor Analytics 

Sensor Analytics 

Person entering 

restricted tunnel 

Suspicious Activity Detection ~ 1 sec 

Pub/Sub 

Client 

Messaging 

Hub 

DBMS 

Pub/Sub 

Client 

Messaging 

Hub 

Abandoned bag 

on platform 

Event Anomaly 

Detection 

Event Anomaly 

Detection 

29 

Correlation of 

Suspicious 

Activity ~ 20 sec 

Distributed Event Miner 

Alert: Potentially Coordinated Attack 




Tracking customer shopping patterns 

How do customers navigate the store? 

● Use Digital Video to establish customer 

traffic patterns 

● Map statistically with any parameters 

• Time of day, day of week, by type of sale 

● Integrate with Dynamic Merchandizing 

• “If I show this content – people go there” 

● Measure customer traffic by product mix 

● Integrate with people count capabilities 

30 




Traffic Flow and Display Effectiveness using heat maps 

Plot all tracks: 

‣ Density 

‣ Flow 

31 




Streaming Data Example: Soccer 

Events 

Informix 

Real-time 

Loader 

(RTL/DSE) 

In-Memory 

Database 

Informix 

Dynamic 

Server 

SQL 

queries 

• Ball, players, and referees are RF tagged (26 transmitters) 

• Position and speed data are streamed to RTL/DSE (± 1.5 cm, 100K messages/s) 

• RTL/DSE stores time-stamped data in database at the rate of 7K-12K messages/sec 

• Prototyped for use in World Cup Soccer 2006 

Periodic 

writes to 

database 

DB2 

Timestamped 

data 

history 

32 




Data Being Captured at Increasing 

Spatiotemporal Resolution 

● People sensors: location 

(inferred from that of their 

devices), activities (inferred from 

calendar and desktop 

information), biometrics, etc. 

● Place sensors: room status 

(inferred from anonymous 

motion/sound detectors), 

presence (of people and things), 

congestion (inferred from 

pressure pads or camera 

images), etc. 

● Thing sensors: location (RFID), 

status (monitoring sensors such 

as telematics, desktop), etc. 

● Business sensors: context from 

databases (medical data, credit 

history, location history), context 

from processes 

33 




General Framework for Time-Dependent Event Handling 

Financial 

Market Data 

RFID Data 

Rich Media 

An event-based 

engine routes, 

sequences, and 

filters event data in a 

time-dependent 

fashion 

Continuously 

Analyze Event 

Streams 

Event Analysis 

Event-based Engine 

Selectively 

Notify Systems 

& Personnel 

Event Notification 

Identify Patterns 

of Interest 

Event Action 

Surveillance 

TD 

DB 

DBMS 

DB 

Data 

Warehouse 

TD: Time-dependent 

34 




Reference Architecture for Event-Driven Middleware 

Realtime SOA 

Data 

Source 1 

DB Tradeoffs for Event-Handling 

1. Latency for Consistency 

2. Throughput for Persistence 

… 

Data 

Source N 

Intelligent, Timedependent, 

Pub/Sub, and 

Routing Hub 

DBMS 

Short-term 

storage 

ETL 

Data 

Warehouse 

Longterm 

storage 

ESB responsible for: 

App 1 

… 

App N 

• High-throughput data handling 

• Low-latency messaging and routing 

35 




Requirements for Event-Driven Applications 

Responsiveness 

Event 

Throughput 

(events/sec 

/server) 

Event Processing Language 

Richness Ease of Use 

Scalability 

Hard real-time 

(deterministic, us) 

100,000’s 

Inductive reasoning 

- Untrained patterns 

- Trained patterns 

Internet scale: 

100,000’s endpoints 

Increasing Capability 

Soft real-time 

(scheduled, ms) 

Near real-time 

(< sec) 

Transactional 

OLTP 

10,000’s 

1000’s 

100’s 

Integration with processes, 

workflows 

General multi-stream 

pattern specifications 

Tools for integrating 

content behavior models 

Tools for distributed 

deployment 

Tools for designing event 

flow 

Collaborating domains 

Managed ESB with event 

services 

Event server clusters 

Data Mining 

OLAP 

10’s 

Sequences, thresholds, 

groups 

Simple event pattern tool 

support 

Single server 

Data Warehouse 

1’s 

Message at a time 

filter/route 

36 



Middleware for Time-Dependent Internet Traffic 



Event 

Throughput 

(events/sec 

/server) 



Scalability 



100,000’s 










(< sec) 

Transactional 

OLTP 

10,000’s 

1000’s 

100’s 

Internet Traffic 


workflows 






deployment 


flow 



services 


Data Mining 

OLAP 

10’s 


groups 


support 

Single server 


1’s 


filter/route 

37 




Middleware for RFID Applications 


Event 

Throughput 

(events/sec 

/server) 



Scalability 



100,000’s 










(< sec) 

Transactional 

OLTP 

10,000’s 

1000’s 

100’s 


workflows 





RFID for retail, distribution, 

manufacturing 


deployment 


flow 



services 


Data Mining 

OLAP 

10’s 


groups 


support 

Single server 


1’s 


filter/route 

38 




Middleware for Surveillance Applications 


Event 

Throughput 

(events/sec 

/server) 



Scalability 



100,000’s 










(< sec) 

Transactional 

OLTP 

10,000’s 

1000’s 

100’s 


workflows 





Surveillance Markets 


deployment 


flow 



services 


Data Mining 

OLAP 

10’s 


groups 


support 

Single server 


1’s 


filter/route 

39 




Middleware for Financial Services 


Event 

Throughput 

(events/sec 

/server) 



Scalability 



100,000’s 










(< sec) 

Transactional 

OLTP 

10,000’s 

1000’s 

100’s 


workflows 





Financial market information and 

program trading 


deployment 


flow 



services 


Data Mining 

OLAP 

10’s 


groups 


support 

Single server 


1’s 


filter/route 

40 




Industry Implications 

● The real world is being captured (through sensors) and modeled at 

increasing spatiotemporal resolution 

● On demand businesses will need to evolve to deal with (monitor, 

process, store, and respond) the increasingly event-driven world 

● New middleware, programming models, and tools will emerge to 

deal with the increasing volume and time-dependent requirements of 

event-driven data 

● SOA will be augmented with event orientation and processing, 

resulting in a hybrid between a web service and an event 

programming model 

41 




Towards 3-D Internet 

Level of Interaction 

Connected 

Inside 

Individual 

42 

Time 




Early Virtual Worlds Business Applications 

Commerce 

Collaboration 

and Events 

Education 

and Training 

Other 

Emerging 

Applications 

43 




Building an On Demand Environment 

Evolving priorities 

44 




What about today‘s IT Challenges 

IDC Directions 2007, 

CIO Strategies to Build the Next Generation Data Center, 

Doc #DR2007_5VT, Feb 2007 

Spending 

(US$B) 

$300 

$250 

$200 

$150 

$100 

$50 

Power and cooling costs 

Server mgmt. and admin. costs 

New server spending 

65 

60 

55 

50 

45 

40 

35 

30 

25 

20 

15 

10 

5 

Millions 

Logical 

server installed 

base: 

Physical 

server installed 

base: 

$0 

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 

0 

45 




What about today‘s IT Challenges 

● Management costs grew 7X as fast as IT spend over last decade 

● Power and Cooling Costs soon to equal Server Spending 

● 15 years ago 75% of IT budgets were spent on new hardware and 

software; today 75% is spent on just running the infrastructure 

● Complexity and administrative cost are growing faster than 

hardware cost 

● Growing revenue while containing cost remains #1 concern of 

management 

● Businesses cannot address the need for integration of people, etc 

without first addressing the underlying infrastructure 

46 




Event Driven IT 

Requires an Event Driven Operating Environment 

● Open 

● Virtualization 

● Autonomic 

● Green Technology 

47 




Evolution to an Open Standard 

Need 

Initiator Core group Standards body 

Open 

Customer need for 

technical solution to 

known problem 

Lack of industry 

accepted technical 

solution 

May be competing 

technical approaches 

or single proprietary 

solution 

Lack of 

interoperability 

A company, individual 

or group of 

companies or 

individuals agree to 

address issue 

Resources devoted to 

developing best 

technical solution, 

often in collaborative 

fashion 

Interested parties publish 

specifications 

Specifications publicly 

available sufficient to 

enable implementation, 


Can be implemented 

with little or no 

restrictions 

Developers may create 

reference or commercial 

implementation 

Developers declare 

intent to have solution 

accepted as standard 

Standards body reviews 

technical solution, adopts 

as standard 

Specifications publicly 

available are sufficient to 

enable implementation, 


Can be implemented with 

little or no restrictions 

Standards body open to 

broad participation, open 

decision making process 

Standard implemented in 

competing IT products by 

multiple vendors. 

48 




Spectrum of open standards 

Global 

SMB 

Computer 

Services 

Insurance 

Financial 

Markets 

Accounting 

Banking 

Finance 

Open OS 

Industry-Specific 

Linux 

Business 

Function 

Open Networks 

IP 

HTML 

XHTML 

Energy & 

Utilities 

Human 

Resources 

Open 

Publishing 

Chemicals & 

Petroleum 

Industrial 

Products 

Marketing 

Aerospace 

& Defense 

Automotive 

Telecommunications 

Pervasive 

& Wireless 

Media & 

Entertainment 

Education 

Technology 

Information 


Government 

US Federal 

Open 

Interchange 

Grid 

Procurement 

Life 

Sciences 

Web 

Services 

STANDARDS 

J2EE 

Open Programming 

Standards 

Standards 

XML 

Autonomic 

Open Data 

Customer 

Support 

Standard 

s 

Relationship 


Travel & 

Transport 

Trading Partner 

Collaboration 

Product 

Life Cycle 

Wholesale 

Electronics 

Consumer 

Products 

Retail 

49 




“Event Driven” IT Standards Are Emerging 

Existing and evolving standards 

• Real-time Specification for Java (JSR 1), Distributed Real-Time Specification (JSR 50) 

• Web Services Notification (WS-Notification); Web Services Eventing (WS-Eventing) 

– Standard for describing event distribution mechanism (like pub/sub), WS-Notification is broader and 

includes decoupled (brokered) operation 

• OMG Data Distribution Service (DDS): 

– Standard interface for Pub/Sub communications for real-time systems 

– Has appeared in requirements from many federal RFPs and RFQs 

• IEEE 1516 (formerly, High Level Architecture – HLA) 

– Standard for communications between computer simulations across disparate computing platforms 

– Being explored for distributed simulation in games, life sciences, and communication networks 

Gaps in standards 

• New event programming languages emerging in open source community 

• Web Services and SOA extensions to handle event-driven applications 

• Streaming data standards for analytics and interchange (e.g., “RSS for events”) 

50 




Virtualization 

“Virtualization is the process of 

presenting computing resources in 

ways that users and applications can 

easily get value out of them, rather 

than presenting them in a way 

dictated by their implementation, 

geographic location, or physical 

packaging. In other words, it provides 

a logical rather than physical view of 

data, computing power, storage 

capacity, and other resources.” user 

Jonathan Eunice, Illuminata Inc. 

user 

user 

manufacturing 

Resource Pool 

data center 

design 

user 

partners 

51 





Virtualization can help improve the Utilization of 

IT Resources, Information and Human Assets 

Pools of virtual resources 

Servers Storage 

Distributed Network 

Partitioning 

SAN volumes 

GRID 

VLANs 

Clustering 

Workload 


TotalStorage 


TotalStorage 


expanded capabilities 

Server allocation for 

Web application 

servers 

ISV Grid middleware 

HiperSockets / 

Virtual ethernet 

Differentiated 

services 

Vendor alliances 

52 




Virtualization Layers 

Applications 

Virtual Server 

Hosting 

Heterogeneous 

Workload 


End-to-End 

Automation 

Middleware 

Shared 

Services 

Security 


End-to-End 

Systems 


Disaster 

Recovery 

Heterogeneous 

Clustering 

OS OS OS 

Partition 

Hypervisor 

Hardware 

Dynamic LPAR 

CIM 

Lightpath I/O 

Software Rejuvenation 

Virtual LANs 

Wireless Support 

53 




Server Virtualization: Dynamic LPAR 

200 

200 

OS OS OS 

150 

100 

Whitespace 

System B 

Whitespace 

System A 

150 

100 

Whitespace 

System B 

System A 

Partition 

Hypervisor 

50 

0 

50 

0 

Hardware 

08.00 

11.00 

14.00 

17.00 

20.00 

23.00 

02.00 

05.00 

08.00 

11.00 

14.00 

17.00 

20.00 

23.00 

02.00 

05.00 

OS1 OS2 OS3 

+WLM 

Processors 

P P P P P P P P P P P P P P P P P P P P P P P P 

Memory 

Memory 

Memory 

Memory 

Memory 

Memory 

Channels 

I/O 

I/O 

I/O 

54 




Spectrum of Server Virtualization 

Resource virtualization and application virtualization are complementary and can co-exist 

App 

App 

App 

App 

App 

App 

LPARS, 

PPARS 

Resource 

Partitions 

Runtime 

Environment 

OS 1 

Hypervisor 

Hardware 

Server 


Runtime 

Environment 

OS 2 

PHYP, 

Xen, 

VMware 

Runtime 

Environment 

Mediator 

OS 

Hardware 

Runtime 

Environment 

Application Environment 


PDS, Softricity, 

Meiosys, 

Veritas Appscale 

Java Virtual 

Machines 

CICS Regions 

Deep 

Complete isolation between OSs 

Run different OSs 

Smaller trusted computing base 

Less 

Only HW resources are shared 

Larger container 

More context switching overhead 

Isolation 

Sharing 

Isolation only between Apps 

Running 1 version of OS 

Trust depends on entire OS 

Shallow 

More 

Apps share common OS services 

Smaller container 

Less context switching overhead 

55 




Server and SW Virtualization Is Becoming Pervasive 

● Hypervisors will virtualize the underlying HW platform 

• IBM LPARs, EMC VMware, Xen, HP VM hypervisor (12/2005), MS Virtual Server 

• Future HW support: Intel (VT-x, Oct. 2005; VT-i, 2006) and AMD (Pacifica, 1H06) 

• Hypervisors will become integrated as a BIOS/firmware function in 1-2 years 

● Hypervisors will become bundled with operating Systems 

• MS Virtual Partitions, Sun Solaris containers, HP-UX Secure Resource Partitions 

● Self-virtualizing I/O adapters will become available 

• Fibre Channel N_Port ID Virtualization (NPIV) is a first step 

• Standard PCI-E adapters will become self-virtualizing in 2-3 years 

● Application containers will gain acceptance as an efficient 

method for consolidating many applications on the same OS, 

while providing isolation 

• AIX Corrals leveraging Meiosys, PDS (on Linux & Windows), Veritas 

56 




Next Generation Internet: GRID Computing 

● Shared resources accessed via the Internet or private intranets using open 

protocols 

● Open Source standards coordinated by Globus and the Global Grid Forum 

● Technical foundation for eSourcing, or ‘utility’ computing 

● Grid Computing: Delivering Software, Server and Data Storage Resources 

like Power from a Socket 

57 




Case Study #1: IBM Download GRID 

58 




Case Study #1: IBM Download GRID 

Download Clients 

● Java Application 

• Interfaces directly with the 

download service 

● Java Applet 

• Interfaces with the the download 

service throuh a web application 

• The web applications handles 

web services calls 

• Compatible with every browser 

running Java 

59 




Case Study #2: World Community Grid 

60 





www.worldcommunitygrid.org 

IBM Philanthropic Initiative 

● Mission is to create the largest public computing grid of PCs benefiting 

humanity. 

How It Works 

● Volunteers download a small, free agent at www.worldcommunitygrid.org. When 

idle, PCs request data from World Community Grid’s server, perform 

computations, send the results back to the server and prompt it for new work. 

Human Proteome Folding Project 

● First project will provide scientists with data that predicts the shape of human 

proteins. With an understanding of how each protein affects human health, 

scientists can develop new cures for human diseases. 

Advisory Board 

● International group of experts specializing in health sciences, technology, and 

philanthropy. 

Volunteer Base 

● Individuals, companies, associations, universities, foundations 

61 





62 





63 




GRID Implementierungen 

Inter-Grids 

Inter-Grids 

Extra-Grids 

Extra-Grids 

Intra-Grids 

Intra-Grids 

VPN 

64 




Virtualization Solutions - Stages 

Orchestrate Infrastructure: 

Sense and respond to changes 

based on business policies 

Secure Cross Enterprises: 

Enable internal and external 

integration and resources. 

Automate Workflows: 

Tasks like change/ 

configuration, 

ITIL processes 

Virtualize Outside The Enterprise: 

Suppliers, partners, customers 

and external resources 

(opportunity: Deep Computing Capacity on 

demand centers or ASP solutions) 

Virtualize The Enterprise: 

Enterprise wide Grids and Global 

Fabrics 

(opportunity: Clash Analysis solutions on grid; 

visualization solutions) 

Virtualize Unlike Resources: 

Heterogeneous systems, application 

based Grids and networks 

(opportunity: EWLM for managing heterogeneous 

environment; IDI for MySAP) 

Virtualize Like Resources: 

Homogenous systems, storage and networks 

(opportunity: consolidate similar servers onto a Blades or other 

series machines; Storage virtualization) 

Updated IT Governance and Management Processes 

65 




Built on Standards 

Standards and Open Source 

Blade.org 

66 




Virtualization Platform 

Workload Virtualization 

Provisioning Orchestration 

Resource Pooling Scheduling 

Information Virtualization 

Federation Cleansing 

Transformation Global File System 

Standards and Open Source 

Virtual Access and Management 

Workload Management Usage and Accounting Resource Management Discovery and Mapping 

Resource Virtualizers 

Hypervisors Containers Block Virtualization VIO VLAN 

Servers 

Storage 

Networks 

67 




Value of a Virtualized Infrastructure 

• Improve TCO 

Decrease management cost 

Increase asset utilization 

Decrease Power and Cooling 

Link infrastructure performance to business goals 

• Access Through Shared Infrastructure 

Leverage common tools across many systems 

Improve business resilience and security 

Establish Service Oriented Infrastructure Foundation 

• Increase Flexibility 

Create pools of system resources 

Maintain freedom of choice with open standards 

Simplify by masking complexity 

68 




Autonomic Computing 

IBM Development Initiative for Self-Managing Systems 

Improved Availability and Security 

Lower Costs for Maintaining and 

Deploying Systems 

Self-Configuring 

Self-Healing 

Self-Protecting 

Self-Optimizing 

Self-managing Systems reduce Downtime, 

Staffing and other administrative Expenses 

Today: 

Cost of skills: 

For every $1 U.S. spent on server HW, $10 U.S. 

will be spent on people to manage the HW 

Cost of downtime: 

$10,000 U.S. per minute 

for e-commerce 

69 




Evolution to Autonomic Computing 

Basic 

Level 1 

Rely on reports, 

product and 

manual actions 

to manage IT 

components 

Managed 

Level 2 


software in place 

to provide 

facilitation and 

automation of IT 

tasks 

Benefits 

Greater system 

awareness 

Improved 

productivity 

Predictive 

Level 3 

Individual 

components and 

systems 

management tools 

able to analyze 

changes and 

recommend 

actions 

Benefits 

Reduced 

dependency on 

deep skills 

Faster/better 

decision making 

Adaptive 

Level 4 

IT components 

collectively able 

to monitor, 

analyze and take 

action with 

minimal human 

intervention 

Benefits 

Balanced 

human/system 

interaction 

IT agility and 

resiliency 

Autonomic 

Level 5 

IT components 

collectively and 

automatically 

managed by 

business rules 

and policies 

Benefits 

Business policy 

drives IT 

management 

Business agility 

and resiliency 

Manual 

Autonomic 

70 




Solution Positioning 

Software Policy 


Patch 


Data Center Automation 

LPAR 

VMware 

Physical 

Servers 

UNIQUE REQUIREMENTS 

Highest Value from Automation Solutions 

• Execute repetitive, error-free best 

practices 

Interfaces to Data Center devices 

• Server provisioning 

• Network provisioning 

• Storage provisioning 

• Security provisioning 

COMMON REQUIREMENTS 

Software 

Deployment and 

Packaging 

Software Distribution 

and Management 

(Branch Office – Departmental) 

Servers 

Clients 


Highest Value from Management 

Solutions 

Interfaces to remote branch devices 

Network Considerations 

Scalability Requirements 

• Thousands of autonomous remote 

sites 

Checkpoint Restart at Job/Network levels 

Asset 

Inventory 

Bare-Metal 

Installation 

Software Distribution 

and Management 

(Pervasive – Mobile Client) 

Clients - Pervasive 


Interfaces to What’s Remote 

• Windows 2000, Windows XP, Windows 

CE, Palm OS, phones 

Significantly higher scalability 

requirements 

• Tens of thousands of “intermittently 

connected” clients 

Server control point at central site 

71 




Pervasive Virtualization drives 

IT Management Simplification 

● Redistributing workloads to optimize performance/QoS, 

improved resource utilization 

● Security and isolation through partitioning 

● Provisioning new systems using pre-built images (freezedried 

stacks) 

● Concurrently supporting incompatible environments 

● Robust retention compliance through archive of entire system 

image (data, application, middleware, OS, etc.) 

72 




73

BWI - ETH ZÃ¼rich - Information Management - ETH ZÃ¼rich

Create successful ePaper yourself

Delete template?

Save as template?