home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!olivea!sgigate!sgi!rhyolite!vjs
- From: vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
- Newsgroups: comp.dcom.cell-relay
- Subject: Re: Comment made at Next Generation Networks conference
- Message-ID: <snq1i9s@rhyolite.wpd.sgi.com>
- Date: 23 Nov 92 16:42:12 GMT
- References: <skofpgc@sgi.sgi.com> <By3Lru.BxE@usenet.ucs.indiana.edu> <By5E8o.G0K@usenet.ucs.indiana.edu>
- Organization: Silicon Graphics, Inc. Mountain View, CA
- Lines: 70
-
- In article <By5E8o.G0K@usenet.ucs.indiana.edu>, robelr@bronze.ucs.indiana.edu (Allen Robel) writes:
- > In <smji130@rhyolite.wpd.sgi.com> vjs@rhyolite.wpd.sgi.com (Vernon Schryver) writes:
- >
- > >> Nope. On page II-27 of the spiralbound (November 16th) handout, top
- > >> slide, second bullet item he states:
- > >>
- > >> "100 Mbs ATM delivers Less Bandwidth Than 100 Mbs FDDI"
- > >
- > > Where is the conflict? Rob wrote "this is slightly lower (6.8%) than
- > >FDDI's 97.3 Mb/s", refering to the number of bits a single TCP/IP
- > >virtual circuit might get out of either FDDI or ATM.
- >
- > You just mentioned where the conflict is when you say "a SINGLE
- > TCP/IP virtual circuit." [emphasis mine]. What would be the
- > throughput for that SINGLE circuit when you've got 20 other
- > fast machines using the same bandwidth? Now compare that figure
- > to the same twenty machines connected with an ATM switch...
-
-
- hmmph! Let's not chop nits. I was only trying to point out that there
- exists an interpretation of the quote that is entirely true, and that
- Rob Warnock's numbers were consistent with that interpretation, as well
- as suggesting ways in which an ATM-LAN can deliver more bandwidth.
-
- Obviously, if all 22 machines are bashing the same FDDI ring, then
- an ATM switch might do better. I'd generally expect 11 pairs of machines
- each wanting 100Mb to do better on an ATM switch than on one FDDI ring.
- However, that is not necessarily the case:
- -it might be that the FDDI links go to a hub containing a crossbar
- switch, similar to the ethernet switches, instead of to a
- concenterator.
-
- -what is the nature of the ATM switch? Aren't there some 100 Mb
- switches that are not the equivalent of a full cross bar,
- so that there is some potential for conflict?
-
- I think this part of this thread is uninteresting. The basis seems to
- be a comparison of FDDI and ATM-as-a-LAN. Bandwidth matters, and since
- 155 is more than 100, ATM switches should provide more bits/sec than
- FDDI rings. However, that is not now an major consideration when
- comparing FDDI and ATM-LAN's. More interesting are:
-
- -current ATM adapters tend to (or all do?) require host programed I/O.
- In other words, current ATM products are sloooooooooow! One
- can reasonably hope that will soon change.
-
- -I understand that all current ATM-LAN's use fixed and permanent
- curcuits among all hosts. That seems unlikely to work in a
- real LAN of hundreds of TCP/IP hosts, all coming and going.
- One obvious answer, to treat ATM virtual circuits as is done
- with IP over x.25, does not sound inviting to me.
- Having ARP imply creating a virtual circuit seems strained.
-
- -multicast needs work. The various mechanisms that are currently
- used to implement broadcasts are not sufficient for multicast.
-
- -The Standards Process must be navigated, and as shown by FDDI
- and ISDN, that can take literally decades.
-
-
- ATM switches as LANs offer a lot of hope. If you must now pick
- something that will replace FDDI, then point at ATM. However, many
- years must pass before anyone can know if that hope is reasonable. My
- favorite example of a technology that everyone claimed would solve
- everyone's problems was "plated wire." If you don't recognize it, then
- think of "bubble memory", which had a similar story about 15 years
- later.
-
-
- Vernon Schryver, vjs@sgi.coml
-