Chicago airport, 4pm. I slide a Visa card carefully into the payphone, wait while it touch-tones my account number to the waiting exchange. Please select your carrier. It’s a few years since I called long-distance from a public phone in the US, and deregulation plus competition now dictates that callers should be able to choose who gets their business. Should it be AT&T Co? MCI Communications Corp? US Sprint Communications Co? After an eight-hour flight, do I care, and anyway how do I select it? Six hours of jet lag and the real implications of competition are getting through to me. Eventually I get a verbal list, in alphabetical order, from the operator, and dial 10288 for a route through AT&T. Perhaps there is something to be said for monopolies after all. This free competition is a prime reason for the explosion of wide-area communications in North America, and it’s natural that the industry’s prime exhibition, Interop, should be a hit.
Open Standards
That’s what it was, although some people I met suggested that Interop is a victim of its own success: the best element of the show was its marketing and organisation, faultless, slick and very American. The amount of equipment on show is best illustrated by the fact that next year’s show will be in San Francisco and not San Jose, since the San Jose Conference Center cannot supply enough electrical power to the show floors. A little thin on new products, the principal theme of the show was progress on existing standards. There was an almost complete absence of ISO-related products, and instead a massive presence of de-facto standards like TCP/IP and SNMP, giving the impression of an industry trying to show that things were working, today – and that they couldn’t wait around while standards bodies dithered.It’s comforting that much of OSI is based around TCP/IP, and that the Internet Engineering Task Force (which drives TCP/IP) states that it will move towards open standards as they become available. At this rate, neither party will have to move very far. As predicted, there was a lot of excitement about Frame Relay – centering on its use for local-net-to-local-net interconnection with most router and bridge manufacturers pushing Frame Relay interfaces. Thankfully this becomes relevant in the UK right about now, as British Telecommunications Plc launches ExpressLane, its pilot Frame Relay service. But Frame Relay is unsuited to voice traffic, and US vendors are looking beyond it to the Broadband ISDN technologies that will finally integrate data, voice and video over public, cell-switched networks – ATM or Asynchronous Transfer Mode, the same technology that was a hit in Geneva recently. Switched MultiMegabit Data Services, SMDS, wasn’t so obvious at the European show, it is already being offered by some US carriers – think of it as a giant local area network, spanning the country, with your access line into it running at 2Mbps. Charges are for the amount of data sent, and other sites could be other offices of your company, or perhaps the disaster recovery service you’ve a contract with. If you’ve wondered where the 802.6 Metropolitan Area Networks standard was going, this is it. On the product front, Retix Inc’s venture into Token Ring bridging was on show. While its Ethernet product has just been enhanced with data compression facilities, the Token Ring box is based on an Intel processor and not the Ethernet family’s 68000 – it sounds as though the iAPX-86 line will be the system of choice for Retix in fu-ture systems and we may see the Ethernet range gradually moving to it.
By Richard Thomas
With local networks interconnected by such boxes fast becoming the favourite way for companies to build multi-site networks, many companies are wondering if they can avoid renting separate lines for traditional IBM SNA-based network equipment by connecting them across these backbones. Initial packages, known as SDLC Passthrough from Cisco Systems Inc and Wellfleet Communications Inc, take SNA frames and encapsulate them as TCP/IP or Source Routing packets before throwing the
m into the backbone – these frames can then be routed anywhere in the network and emerge, reconstructed, on another site. While Proteon Inc, Tricom Computer Pty Ltd and other vendors (including IBM itself) have since jumped on the passthrough bandwagon, I saw at least two vendors providing dedicated SDLC-LAN bridges. Vitalink Communications Corp and Sync Research both showed units that convert SDLC traffic from cluster controllers into Logical Layer Control frames (low-level OSI sublayer frames) on a Token Ring. These frames can then travel the backbone and, rather than come back out of a similar box as SDLC the other end, pass straight into a local-network-connected mainframe, via a front-end processor with a Token Ring interface, for example. I suspect we’ll see more of this type of interconnection. Although Network Management was one of the show’s favourite topics, it was still difficult to get demonstrations of multi-vendor management from a single workstation. Every vendor has its own Manager station, usually based on a Unix workstation and which can draw topological maps of your network, colouring segments according to load or status. Click on the router icon and the statistics or management data for that router pop into a window on-screen. Ask to click on a box that isn’t made by the management station supplier, though, and the smiles fade on all but a few stands. Simply saying that a device supports SNMP (most do) isn’t enough to let other peoples’ management stations control that device. Instead, formal arrangements to exchange MIBs – Management Information Bases, definitions of the things which can be managed or examined in a given piece of equipment – must be set up between the different vendors. In this respect Hewlett-Packard Co’s OpenView and Sun Microsystems Inc’s SunConnect are ahead of most. Sun, as an example, has arrangements to integrate management with 3Com Corp, BICC Data Networks Ltd, Bytex Corp, Chipcom Corp, Microcom Inc, Retix Inc and Synoptics Communications Inc. Sadly, offerings that run on personal computers are rare and so users looking for an entry-type management facility to control their fledgling network will be disappointed. OpenView is supposed to run on an 80386-based system, but salesmen hastily steer you away from the machine muttering about low performance and restricted function. Another trend is Ethernet clusters – units that link a group of Ethernet segments with either very high-speed bridges (from Alantec Inc) or switches (Kalpana). By passing packets at full bus speed between segments, while isolating traffic that needs only to pass down a single segment, these devices effectively multiply the bandwidth of an Ethernet.
Saturated Ethernets
The obvious applications are where fast workstations are starting to saturate Ethernet in sites where progression to fibre is too expensive or physically impractical. The Alantec hub bridges up to 12 Ethernets at 60,000 packets per second, and has an FDDI port if you need it. But perhaps the most unusual item came from Chipcom Corp in the form of its Token-Ring-on-a-card concentrator. This full-length AT adaptor contains a conventional 802.5 Token Ring node and an eight-port media access unit and fitting a hydra – fan-out – cable to the multi-way connector on the back of the card creates a complete Token Ring Network with no external components – Ring-in and Ring-out are provided. It’s been developed to take advantage of Novell Inc’s new hub management interface, and currently only has drivers for NetWare. After the last few years of promises, Interop was a relief. It gave the overwhelming impression of an industry that can see light at the end of the tunnel – the shackles of grindingly slow or prohibitively expensive wide-area services are about to be hacked free.