A user recently asked independent research organisation the Yankee Group Is ATM really going to be the foundation of all my networks and is it going to last me until the year 2000, or are you going to spring something new on me next year that makes all my Asynchronous Transfer Mode equipment obsolete? The answer Yankee Group gave was ATM is it. And this, broadly, is the conclusion of its latest report, The Future of LAN Switching. The case for Asynchronous Transfer Mode rests largely on its technical and potential cost saving advantages over other options, says Yankee Group. Not only does it offer a protocol that will be employed in both the wide area and the local network, simplifying network management as well as provisioning and lowering network costs, but, operating at Layer One and part of Layer Two of the Open Systems Interconnection stack, Asynchronous Transfer Mode is a simpler protocol than current local networking protocols and a simpler switching technology than the packet switching employed in Ethernet, Token Ring, or FDDI switching. This will make Asynchronous Transfer less expensive to implement than these other technologies. In 1992, says the report, Asynchronous Transfer Mode products were introduced by vendors such as Adaptive Corp and Fore Systems Inc with per-Network Interface Controller and per hub port costs of $4,000 and up.
ATM Forum
This year Yankee Group anticipates that vendors will introduce Asynchronous Transfer Mode technology for $1,000 to $1,500 per hub port and expects costs to drop even further to $500 by 1996. Because it has no MAC Media Access Control layer functionality, Asynchronous Transfer Mode is not tied to a specific speed or medium, unlike Ethernet, Token Ring and FDDI. With ATM, users will not have to put up with lengthy standards development timelines and costly upgrades to go from fibre to twisted pair or from 10Mbps to 100Mbps, the report continues. Asynchronous Transfer Mode addressing will enable users to set up logical local networks comprising users on different local network segments or different local networks easily . It will be able to establish the relationships between desktop devices on the network, and will also be able to react to changes in the network without involving higher layer ( – Layers Two and Three) protocols or requiring the network administrator to establish and maintain complex relational databases. Asynchronous Transfer Mode will scale into very high speeds in the optical carrier range of OC12 – 622Mbps, OC48 – 2.5Gbps, and up. So what is keeping Asynchronous Transfer Mode from taking over the world? It is simply not here yet, finds Yankee. Although the ATM Forum has been instrumental in accelerating the development of Asynchronous Transfer Mode standards, in reality it has only scratched the surface in defining the set of interfaces and protocols that will be needed before full scale Asynchronous Transfer Mode networks can be constructed. With current ATM standards, there are several factors that will dissuade users from investing in ATM before late 1994. These include: no support for switched virtual circuit and no switched virtual standard; no finalised addressing standard; multicast support issues; the current limited speed and media choices; and lack of a public and private network-to-network-interface standard.
By Ian Holland
The vendors aiming to bridge these gaps are coming from the local network and wide area arenas, with each seeing an opportunity to cross into the other environment. Asynchronous Transfer Mode products or statements of direction have come from all the dominant wide area network customer premises equipment vendors, such as StrataCom Inc, Newbridge Networks Inc, Network Equipment Technologies Inc’s Adaptive, Ascom Timeplex AG, Bolt Beranek & Newman Inc, and Netrix Corp. In addition, Asynchronous Transfer Mode local network-targeted products or product announcements have emerged from Adaptive, Bolt, Beranek – with Ungermann-Bass Inc, and Newbridge. These vendors face the challenge of no installed base in the loca
l network and little experience in local network interfaces. However, they bring excellent network management and network design expertise, experience with complex and mission-critical networks, and a foot in the door with the central data processing department for the provisioning of the wide area network. Wide area vendors are forming alliances with local network router vendors to obtain the needed expertise in local network software. Yankee expects these to be a force in the Asynchronous Transfer Mode local network-switching market, with their presence most keenly felt in the enterprise switch market. The router vendors will compete with the wide area switch vendors from the position of enterprise switch. Their strategy focusses on augmenting current high-end routers with a large concentration of local network ports, thus enabling the router to perform as a collapsed backbone, with hubs star-wired into the router. Packet processor interface modules at the port level enable theses routers to switch local network packets at native local network speeds, and Gbps buses enable routing between, for instance, an FDDI and an Ethernet local network to take place at high speed. With the exception of Wellfleet Communications Corp and Network Systems Corp (and Proteon Inc from its hub side) the router vendors have not announced their intention to add Asynchronous Transfer Mode switching capabilities to the routers. Rather they will offer ATM interface capabilities to an ATM switch from other vendors. This strategy build on the strengths of router vendors, which lie in their software. Yankee Group believes this strategy will see router vendors through to the end of the decade, by which time the economics of Asynchronous Transfer Mode, the decreasing need for routing in ATM-based networks, and the trend toward incorporating routing capabilities into other networking equipment, notably hubs, will push routers into a commodity market and prompt router vendors to look elsewhere for growth. Hub vendors are sitting pretty in the local network-switching and Asynchronous Transfer Mode markets with major advantages including their penetration (in terms of installed base) into the local network market; their penetration (in terms of physical placing of hubs) into the corporate wiring structure; their position as focal points for local network management; and their architecture. Hub vendors recognise the opportunity that switching hub technology offers. Their deployment process will follow a four stage evolution. Stage one will be to integrate a multiport bridge-router module in their hubs to establish multiple local network segments with multiple backplanes in the hub. Stage two will see them using a segmented backplane in the hub to establish multiple local network segments, and dedicate bandwidth to each segment. Stage three will involve a switching backbone to establish a virtual packet switching backplane connecting hub ports and/or modules.
Technology partner
Stage four uses a high-speed Asynchronous Transfer Mode switching backplane in the hub as a fabric for bandwidth-on-demand switching. Hub vendors are initially turning to the router vendors as OEM partners for the local network packet-switching component of the stage three hub. However, too much of the manufacturing cost of the hub is in the switching components for the hub vendor to be price-competitive as well as maintaining a healthy profit margin, while giving the router vendors their cut. Because of this, hub vendors are developing their Asynchronous Transfer Mode technology in-house or jointly with a technology partner (Sigma Systems Inc and Optical Data Systems Inc; 3Com Corp and Synernetics Inc; SynOptics Communications Corp, Kalpana Inc and Cisco Systems Inc). A further implication of this OEM-then-home-grown strategy is that hub vendors will be motivated to push their Asynchronous Transfer Mode products (with potentially higher profit margins) over their local network-switching products. The report concludes that users’ and vendors’ plans for ATM in general focus on
protecting legacy systems and wiring infrastructure, providing a means to upgrade first the backbone, then end stations, on an as-needed basis.
