It is difficult to talk to a US West Coast techhie these days without the topic of Asynchronous Transfer Mode communications to the desktop cropping up in the conversation. Most commentators agree that it will be 1993 or 1994 before it bears fruit, but the Retix Inc-Network Equipment Technologies duo beg to differ. They continue to insist that, first, they have signed up customers desperate to build private ATM networks, and second, that they will be ready to ship in October. What we are looking at here though is not Asynchronous Transfer Mode to the desk per se, but ATM to the hub. Both Retix and Ungermann Bass Inc are aiming for the same market with their respective RouterXchange 7000 and Access/One Asynchronous Transfer Mode modules. The market in question is that of ATM concentrators: devices that take slow speed traffic from FDDI, Ethernet or Token Ring local area networks and feed it into the Aysnchronous Transfer network. More details have emerged about the tie-up between Bolt, Beranek & Newman Inc’s BBN Communications and Ungermann Bass announced earlier in the month. BBN started talking about its Emerald Asynchronous Transfer Mode switch last year, just before Telecoms ’91, and the initial specs looked impressive since it is designed have a wide array of interfaces handling speech, image and various types of data. Given Emerald’s ambitions, the tie-up with Ungermann might look a bit curious, after all the BBN device is being designed to incorporate the requisite local area network and voice interfaces, so why the need for Asynchronous Transfer Mode concentrators hanging off the front? In a nutshell, cost. BBN knows very well that Emerald is not the kind of thing that people will want to buy lots of – one per site probably, to be used as a connection between various local and wide area networks. Hence the alliance with the hub manufacturer and the technology agreement that guarantees delivery of traffic to the Emerald switch. Apart from that, Ungermann, apparently, is helping BBN with designing Emerald’s Fibre Digital Data Interface. While the vendors have been planning products, the issue of Asynchronous Transfer Mode standardisation has also been moving apace. Last month saw a letter ballot emerge from the US T1S1.5 Asynchronous Transfer Mode sub-working group committee entitled ‘Broadband ISDN – ATM layer functionality and specifications’. The ballot closed on June 21, and if the proposal is accepted, it should iron out quite a few of the wrinkles in the lower layers of the standard. Meanwhile, rather higher up the layers, a plenary session of the T1S1 committee has unanimously passed the specifications of ATM Adaption Layer type 5, AAL 5. AALs define those services that can be called upon by network applications, and specify to which type of connection any particular Asynchronous Transfer Mode cell belongs. ATM cells consist of a 48-byte payload topped by 5-bytes of header information. AAL data is stored inside the payload, adding to the overhead, although if the information wasn’t added by the AAL system then it would normally have to be added by the application writer anyway. But AAL 5 is special, according to Doug Hunt, senior manager of technical development with BBN. It is the nul type, containing no information. This cuts network overhead, increasing capacity, but moreover, says Hunt, the type cuts the amount of processing that goes on in the system. The result? AAL 5 is useful for high performance applications and interests the likes of Sun Microsystems Inc and Apple Computer Inc.
