home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!europa.asd.contel.com!darwin.sura.net!zaphod.mps.ohio-state.edu!cs.utexas.edu!sun-barr!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: <smji130@rhyolite.wpd.sgi.com>
- Date: 22 Nov 92 18:48:33 GMT
- References: <skofpgc@sgi.sgi.com> <By3Lru.BxE@usenet.ucs.indiana.edu>
- Organization: Silicon Graphics, Inc. Mountain View, CA
- Lines: 36
-
- In article <By3Lru.BxE@usenet.ucs.indiana.edu>, robelr@bronze.ucs.indiana.edu (Allen Robel) writes:
- > ...
- > >No, I fear the eminent McQuillan may have mis-spoken this time. (Or you may
- > >have mis-quoted perhaps?)
- >
- > 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.
-
- I think most applications do not use symmetric flows of data. That
- "100 Mbs ATM delivers" almost two links, each slightly less than one
- "100 Mbs FDDI" link is interesting, but not important for current
- applications.
-
-
- By the way, Rob is not quite right about the ratio of TCP ACK's to TCP
- data segments in 4.3BSD-reno TCP/IP stacks. 2:1 is what you commonly
- see between machines of similar speed on a medium that is not
- saturated. That ratio can be 1:1 if the receiver is very much faster
- than the sender and/or the medium. I have seen the ratio (at steady
- state, moving 10's of GB over FDDI) above 8:1 when the receiver is the
- bottleneck, not the medium or the transmitter.
-
- For example, when the FDDI ring is at 100% utilization (ie. token
- latency at several milliseconds) when an R4000 based Indigo is sending
- to an R3000 based machine, the ratio is just under 3:1, at least on one
- test ring of around 5 usec latency.
-
-
- Vernon Schryver, vjs@sgi.com
-