The competing players in the race to establish a standard for 100Mbps Ethernet have released details of their technologies. On November 9, the various interested parties presented their strategies to the IEEE 802.3 committee, which will convene a Study Group in January to make a decision on a standard. Although most attention has been focused […]
The competing players in the race to establish a standard for 100Mbps Ethernet have released details of their technologies. On November 9, the various interested parties presented their strategies to the IEEE 802.3 committee, which will convene a Study Group in January to make a decision on a standard. Although most attention has been focused on the 100Mbps speed, the proposals actually came under the general heading of high-speed Ethernet, and one company, Cabletron Systems Inc, suggested an implementation of 802.3 that operates in full duplex mode for 20Mbps transmission. The others are all proposing 100Mbps implementations, and fall into two main groups. In the blue corner are Hewlett-Packard Co and AT&T Microelectronics, which have developed a standard called 100Base-VG; and in the red corner are 3Com Corp, Sun Microsystems Inc, SynOptics Communications Inc and LAN Media Corp. Start-up Grand Junction Networks Inc presented a technology essentially similar to the 3Com-LAN Media proposal.
The proposed standards cover much the same ground – support of 100Mbps transmission over existing Ethernet wire. In both cases, network diameter has to be reduced, from around 10,000 feet at 10Mbps to around 750 feet – although this is no smaller than the requirement for 10Base-T networks. But while the Hewlett-AT&T standard requires changes to the Media Access Control layer protocols, the other companies that made presentations claim they can preserve the layer more or less intact. Nigel Oakley, 3Com’s UK product marketing manager, commented, we’re quite early here. Our intention is to make no changes to the MAC layer. But it may be that people come up with problems that mean changes need to be made. Hewlett and AT&T Microelectronics gave the most detailed presentation – as befits the group that wants to make the most changes to existing standards. In fact, while the pair were allotted a two-hour time slot, they ended up speaking for four hours. 100Base-VG, according to the companies, includes full support for all the implementations of 10Base-T wiring specifications, cable distances, cable bundles and connectors and so on. It is also, like its rival proposals, designed specifically for a star topology. According to Mark Raymond, senior analysts at consultancy company DataPro, the signalling overhead involved means that it is mathematically impossible to take a bus-based Ethernet and speed it up while still retaining shared media.
By Emma Woollacott
This means that to upgrade an existing Ethernet network to 100Base-VG, a new high speed network concentrator has to be installed in the wiring closet and connected to the existing backbone wiring. The existing twisted pair cables coming from each computer are unplugged and plugged into the ports on the new concentrator. Secondly, the existing local network adaptors in each workstation are replaced by new high-speed 100Base-VG adaptors; migration can be phased in by individual computers or whole workgroups. 100Base-VG subnetworks can connect to existing 10Mbps Ethernet networks using a 100:10 speed-matching bridge, or to 16Mbps Token Ring backbones or subnetworks, using a router.The proposal is based around two innovations in particular, in the areas of signalling and media access. The original Ethernet bus topology meant that intelligence was needed at each node to manage access to the shared medium. While 10Base-T introduced support for a hub architecture, network access was still managed separately by each node, so that two pairs of wires were needed at each node, one for transmitting and one for receiving. The Hewlett-AT&T Demand Priority Access Method, DPAM, goes one step further to centralise management. A node wishing to transmit first indicates its request to the intelligent hub. Simultaneously, it requests a level of service, either normal, or high priority – which supports high speed applications such as video. But the main advantage of Demand Priority Access Method is that, used with a hub-based architecture, it eliminates the need for an end node to receiv
e at the same time as it transmits. It thus frees up all four wires, which can then be used for tranismission – an ability the companies call Quartet Signalling. Larry Birenbaum, vice-president of engineering at rival company Grand Junction, says Quartet Signalling is impressive, but makes demands on users. From what we’ve seen of it, it has the virtue of appearing to use existing wiring – if you conform to their bundling restrictions. In the US, 25-pair bundles are used. According to modern wiring standards, these are confined to wiring closets, but Hewlett-Packard says many people use them outside, and their scheme is focused on this, he said. These wiring issues are the main reason why the Demand Priority Access Method proposal demands significant changes to the Media Access Control layer.
While this is capable of supporting Category 3 – speech grade wiring with bundles outside the wiring closet, the other two main contenders – Grand Junction and the LAN Media-3Com-SynOptics group – say this is unneccessary. They say they can support Category 5 wiring – which has better electrical characteristics and is used for applications like Fibre Distributed Data Interface – and expect to be able to support Category 3 in cases where bundles are confined to wiring closets. They believe this covers most users’ needs. Datapro’s Raymond predicts that the IEEE’s final decision will combine elements from both camps. 100Base-VG should have a good chance of being the basis of a standard, as Hewlett-Packard’s laboratories were behind a lot of the work on 10Base-T. It was also the most detailed proposal to be set before the IEEE. In any case, believes Raymond, the changes to the Media Access Control layer required by the rival proposals are likely to be more extensive than the companies are trying to make out. It looks as if even the Cabletron 20Mbps proposal would require changes to the MAC layer, if only small ones, he said. The Study Group is expected to make its recommendations by mid-1993, which means commercial products could hit the market before 1994 – putting 100Mbps Ethernet into a similar time frame as the emerging products based on Asynchronous Transfer Mode, and the developing variants of FDDI such as CDDI and FDDI-FOL.