San Jose-based 3Com Corp has been busy in the patents office again. It has logged applications for a high-speed implementation of Ethernet and two low-cost methods of tying isolated work groups local area networks and homeworkers into the corporate network. The company reckons that Ethernet running at 100Mbps will be capable of supporting isochronous traffic […]
San Jose-based 3Com Corp has been busy in the patents office again. It has logged applications for a high-speed implementation of Ethernet and two low-cost methods of tying isolated work groups local area networks and homeworkers into the corporate network. The company reckons that Ethernet running at 100Mbps will be capable of supporting isochronous traffic such as speech or video. The two-pronged high-end and low-end attack on the market was revealed by John Hart, chief technical officer, who was pressing the flesh in London last week. With the rest of the data communications world wrapped up in developing desktop connections based on the Copper Distributed Data Interface or Asynchronous Transfer Mode technology, 3Com’s decision to concentrate on speeding up Ethernet seems to be a somewhat idiosyncratic one.
But Hart argues persuasively for the technology on both technological and business grounds and hotly disputes that the company is throwing a spanner in the Copper works. It is a mistake, he says, to assume that technology admirably suited to wide area links or campus backbones is necessarily suited to the last 100 yards between desktop and hub. Hence the need for yet another 802 series local area network standard – a standard that he believes is very likely to emerge – although the admits that there isn’t any action in the standards committees at the moment. What Hart proposes is a system similar to that used in dual speed 4/16Mbps Token Ring adaptors. Speeding up Ethernet should not prove difficult provided that segment lengths are restricted. The existing speed of 10Mbps is, says Hart, merely a symptom of the distances that it was designed to cover. Design a technology that relies on collision detection technology and has to connect devices between one mile and seven miles apart, and it has to run relatively slowly so that all stations along the cable length can see transmitted packets simultaneously. Restrict the maximum length and Ethernet can run much faster Hart proposes around 100Mbps. 3Com is pinning its hopes on its ability to produce dual-standard adaptor cards at a small premium, perhaps 20% to 40% over the existing price.
By Chris Rose
If it can manage this, then it believes that users will snap them up in the knowledge that they will be able to switch from 10Mbps to 100Mbps without having to tinker with the desktop hardware. In fact, says Hart, the network manager should be able to switch speeds without the end user even noticing. More controversially, Hart says that such a system should be capable of supporting isochronous time-critical data, such as speech and video. The conventional approach to isochronous traffic is to use network protocols that offer guaranteed delivery times. FDDI-2 and Asynchronous Transfer Mode, for example, can reserve slots in the data stream for video or voice, thus ensuring that it isn’t affected by contention. With the levels of compression that most multimedia traffic employs and each desktop device sitting at the end of its own 100Mbps link, Hart reckons that his super-Ethernet will be able to deliver such traffic to the desktop, leaving the Fibre Distributed Data Interfaces and Asynchronous Transfer Modes of this world firmly on the backbone. Meanwhile, at the low end of the local area network world, the 3Com boffins have turned their attention to the plight of the dispossessed souls that either work from home or are stationed on small workgroup networks at remote locations. Both groups frequently need some way to link to the headquarters network and 3Com has developed an approach that it believes will enable it to produce low cost offerings. Taking the workgroup networks first, 3Com looked at existing routers and bridges and decided that the the first were too complex and pricey a way to connect small networks. Also, bridges were felt to be over-engineered and they suffer from multicast data clogging the slow links. Instead, the company has filed a patent on what is essentially a cut-down router with neither the need nor capacity to build complex routin
g tables. Each satellite local area network has a routing adaptor that receives packets destined for elsewhere and passes them over the wide area link to the boundary router attached to headquarters.
It may sound like simple bridging, but Hart points out that the routing adaptor is not operating in promiscuous mode, that is, not examining every packet placed on the workgroup local area network. Rather, like all routers, it deals only with those packets sent to it by the stations on the remote local area network. The result? The routing adaptor doesn’t need the horsepower of a bridge or the complexity of a full blown router and it ought to be cheap – as cheap as a high-end adaptor board, according to Hart, who adds that actual products should appear over the coming 12 months. The idea for single remote users is similar – put a remote hub on the local area network, similar to a terminal server, and then offer a modified network adaptor card that can connect to either a modem, ISDN or any other medium that ties the personal computer into the local area network. So why the delay? The problem is not with the core technology, says Hart, but indicative of the number of wide area interfaces that the company must support.