10.07.2015 Views

DTJ Number 3 September 1987 - Digital Technical Journals

DTJ Number 3 September 1987 - Digital Technical Journals

DTJ Number 3 September 1987 - Digital Technical Journals

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

A <strong>Digital</strong> Network Architecture Overoiewbits per second. The Ethernet has a maximum stationseparation of 2.5 kilometers with a maximumof 1024 stations. A shielded coaxial cableis used as the physical medium. Ethernet alsouses a branching, nonrooted tree topology. TheEthernet IAN technology was jointly developedby <strong>Digital</strong> Equipment Corporation, Intel Corporation,and Xerox Corporation. 18 The Ethernetspecification, with minimal changes, has subsequentlybeen standardized by the IEEE 802 LocalArea Networks Committee as IEEE Standard802.3.19The Ethernet data link protocol provides abest-effort delivery service. Messages, calledframes, are transmitted over the physical chatmelin a broadcast fashion. Stations are assigned48-bit addresses, and each frame contains asource address and a destination address. A framecan be addressed to an individual station or to agroup of stations, using a 48-bit group address(called a multicast address in Ethernet terminology). A special multicast address, consisting of1 's, is used to denote the set of all stations on anEthernet and is typically used for maintenancepurposes. In the DNA architecture, the multicastcapability is used for network configuration purposesby the routing and network managementlayers. For example, a multicast address, specifiedby the architecture, is defined in the routinglayer specification as the set of all routing nodeson an Ethernet IAN.End nodes advenise their availability to routingnodes by periodically broadcasting "hello"messages to the multicast address. The large48-bit address space permits a unique address .tobe assigned to each Ethernet station when it ismanufactured. That address space permits stationsto be plugged in to a IAN and operate withouthaving addresses assigned manually. MOPuses this address when down-line loading computersystems, such as server systems with nomass storage. The 48-bit address can be used toselect the correct program and parameters to beloaded into the node, such as the 16-bit DECnetnode address.The Ethernet data link protocol frames messagesusing the propenies of the Manchester codingscheme employed by the physical channel tomark the beginning and end of each frame. Inaddition to source and destination addresses,frames employ a 16-bit protocol type field toidentify the higher-level protocol carried inthe frame. The protocol type field values areassigned in blocks to all vendors who manufactureEthernets, thus permitting different proprietaryand public protocols to coexist in a singleEthernet station. Ethernet frames also contain a32-bit CRC to ensure that frames received inerror are detected and discarded.Since the Ethernet physical channel can transmitdata only from one station at a time, the Ethernetdata link protocol must allocate the singlechannel among all the stations. This allocation isaccomplished by the technique of CSMA/CD(carrier sense multiple access with collisiondetection) . In this contention-based protocol,stations "listen" before transmitting (carriersense) and defer their transmissions to other stationsalready transmitting.Should several stations begin transmittingsimultaneously, a collision will occur, preventingcorrect reception of any transmission. In thiscase the physical channel hardware in each collidingstation will detect the collision (collisiondetection) and each station will rescheduletransmission after a randomly selected delay. Toensure efficient, stable operation of the networkunder both low- and high-load conditions, thisrandom delay is adjusted on subsequent collisionsby the back-off algorithm. This causes eachstation to reduce the load presented to the networkunder overload conditions. Studies haveshown that this procedure provides good performance(low delay and high throughput) over arange of Ethernet configurations and loads.20•21These studies have also shown that the procedureallocates resources fairly between competingstations and operates stably under high-loadand overload conditions.<strong>Digital</strong> recently introduced the concept ofbridges and extended IANs as a means to extendthe physical extent, number of stations, andthroughput capabilities of a single LAN . 22Extended IANs operate transparently to higherlevelprotocols, such as the.DNA protocols. Thus,although not a part of the DNA architecture,extended LANs - such as those built fromEthernets and LANBridge 1 00s (Ethernet-to­Ethernet) - can be components of a DECnetnetwork.Physical Link LayerThe physical link layer transmits bits of informationbetween adjacent nodes. The functions inthis layer include encoding and decoding signalson the connecting channel, performing clock22<strong>Digital</strong> TecbnlcalJournalNo. 3 <strong>September</strong> 1986

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!