T H E C - G A T E W A Y ------------------------- Purpose: -------- The purpose of this document is to give the reader an overview of the C-Gateway, what it is, what it does, and what the charateristics and status of the different version are. Overview: --------- The C-Gateway was designed by Noel Chiappa as a multi-protocol packet forwarder. It was originally written to run on LSI-11s running under a small operating system written in PDP-11 assembly language from SRI called MOS. It was designed primarily around the idea of fast local area networks but it also works with networks of other design. For instance, we also use it with the ARPAnet, with a network of slow (1200 and 9600 bps) async. serial lines, and with a 4800 bps syncronous line. The Different Versions: ----------------------- There are now three main versions of the C-Gateway. The original version which runs on the LSI-11's is now running on only two machines that I know of and one of those is scheduled for replacement within a couple of days. Noel Chiappa has continued to develop the gateway seperately from the MIT Lab for Computer Science and is marketing it as a product through Proteon Associates, Inc. David Bridgham has replaced MOS and ported the gateway to a 68000 machine and to the uVAX (now maintained by MIT Telecom). - The Original The gateway from which the other two descend is the original one that ran at LCS on 11/03's and 11/23's. The interfaces that it has drivers for are the ProNet ring, the Interlan ethernet, the 1822 Imp interface, a dh lookalike for the async serial lines, the Xerox 3Mbit experimental ethernet, and the Chaos net hardware. There is a `robustness' board which has a timer which will reboot the machine if the code crashes and has roms which contain a TFTP boot program. This boot program has a wired in driver for the ringnet, so a ringnet is required on all our gateways. The only protocol forwarder which is used, and therefore works, is IP. The code does not know how to deal with the 11's memory mapping so only 64 Kbytes is available. The gateway fits in this but there is no room for any more expansion. It will nominally forward 120 packets/sec with an 11/03 and 180 packets/sec with an 11/23. - Proteon (Noel Chiappa) The gateway that Noel Chiappa is selling through Proteon is basically the same gateway. It runs on the same hardware and can use the same networks. However, it does know how to work the mapping hardware and so can have many more, and larger, packet buffers and has more room for expansion of the code. This gateway has a working Chaos forwarder as well as IP. It also has had some performance improvements and can get over 250 packets/sec with an 11/23. Noel has also ported his gateway to a Multibus 68000 machine. This gateway gets significantly better performance than the PDP-11 gateways. I don't know the details of what's available in the way of netbooting and reboot timers or what devices are supported. Although there is at least the Proteon ring, some form of ethernet, and a sync or async serial line. Talk to Noel Chiappa (jnc@xx.lcs.mit.edu) or John Shriver (jas@proteon.com) for more information on these gateways. - MIT Telecom (David Bridgham) The gateway done by David Bridgham was originally a port to a Q-Bus 68000 processor, but this gateway was never made into a production gateway. Soon after, the work was moved to a uVAX-I (and later to the uVAX-II) which has been put into service. It also has a chaos forwarder. Devices that work with this gateway are the Proteon ringnet and the DEQNA. If there is a DEQNA it can use this device to provide a reboot timer. There is netboot code which can be placed in the boot blocks of a disk. Eventually this will be moved onto a ROM card of some sort.