home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!usc!isi.edu!finn
- From: finn@isi.edu (Greg Finn)
- Newsgroups: comp.dcom.cell-relay
- Subject: Re: Future of IP routers
- Message-ID: <22284@venera.isi.edu>
- Date: 28 Aug 92 21:57:31 GMT
- References: <1992Aug27.174448.28143@fokus.gmd.de>
- Sender: news@isi.edu
- Reply-To: finn@dalek.isi.edu (Greg Finn)
- Organization: USC-Information Sciences Institute
- Lines: 25
-
- In article <1992Aug27.174448.28143@fokus.gmd.de> icarus!mbu@fokus.gmd.de (M. Burak) writes:
-
- >Do you really think that TCP/IP over B-ISDN will be the real
- >challenge? As far as I know, TCP/IP allows you a throughput of 28
- >Mbit/s over a 140 Mbit/s network. I don't think that this will be the
- >ultimate protocol performance for high speed networks.
-
- TCP/IP is an internetwork protocol suite. Packets are just
- data. I send packets across my 600 Mb/s network at 400 Mb/s to hosts.
- Operating system architecture, processor speed, and workstation
- hardware architecture are currently the strong limits to
- communications performance and they have little if anything to do with
- protocols. They exist no matter what protocols you pick.
-
- To give you a concrete example, for a typical Sun workstation
- it is difficult to make programmed-loop memory transfer exceed 60
- Mb/s. You find out very quickly when you build high-speed LANs that
- the inefficiency of protocol implementation ranks well down the list,
- behind operating system architecture, interface architecture, memory
- architecture, bus speed, and finally, processor speed. After you fix
- those problems, a particular internet protocol suite may become an
- important limitation, but not before.
- --
- Gregory Finn (310) 822-1511
- Information Sciences Institute, Marina Del Rey, CA 90292
-