home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.dcom.cell-relay
- Path: sparky!uunet!mcsun!dxcern!brian
- From: brian@dxcern.cern.ch (Brian Carpenter CERN-CN)
- Subject: Re: Future of IP routers
- Message-ID: <1992Aug28.155703.4166@dxcern.cern.ch>
- Organization: CERN European Lab for Particle Physics
- References: <1992Aug25.123428.26295@ccsun.strath.ac.uk> <BtKDH9.7M7@usenet.ucs.indiana.edu> <1992Aug26.092945.4663@ccsun.strath.ac.uk> <3445@ra.nrl.navy.mil> <1992Aug27.174448.28143@fokus.gmd.de>
- Date: Fri, 28 Aug 1992 15:57:03 GMT
- Lines: 55
-
- icarus!mbu@fokus.gmd.de (M. Burak) writes:
-
- >Randall Atkinson writes:
-
- >> Claims that one cannot do multimedia over IP look kind of
- >>silly when you have seen real-time voice/video over today's Internet.
- >>Given some tweaking of things, I'm quite sure that better quality and
- >>performance is possible while still using a connectionless network
- >>protocol and doing routing at the network layer.
-
- >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.
- >Isn't it better to use faster and probably new protocols on top of the B-ISDN protocol
- >stack?
-
- The discussion is about IP, not TCP. It could equally well be about
- CLNP, or IPX. IP will limit you to 70% or so of raw ATM bandwidth
- (depending on what assumptions you make about AAL overhead and average
- packet size). Then you have to think about transport overhead, but
- that's a completely independent discussion. Now to get religious:
- if you use AAL3/4 then you can use a very lightweight transport but
- if you use AAL5 you'll need an error-correcting transport such as
- Extended TCP, TP4 or XTP. Current TCP is certainly inadequate, as its
- designers are the first to admit.
-
- >Another point is that I don't understand the whole discussion about IP Routers. Nobody
- >writes that in the B-ISDN there will be Connectionless Servers, taking care of routing
- >*and* security and address screening and all other management related stuff. Or are
- >you talking about setting up connections like for video/audio services using AAL type 5?
- >If the latter is right than the discussion seems for me ok but if you are talking of an
- >public ATM network as envisaged by CCITT then you have the CLS and AAL type 4. Please
- >tell me if that's right or not.
-
- Well, SMDS and CLS already show every sign of being as irrelevant to
- the real world as CONS or ISDN :-). Once I get a private or public
- ATM connection (a 155 Mbit UNI or NNI plug in my modem room), _I_ will
- decide which AAL I use and how I use it.
-
- The argument about routers has been raging also on the IP-over-ATM
- working group list (my fault!). My personal conclusion is that in some
- cases IP hosts will establish end-to-end VCs to peer hosts, and other
- cases they will talk to a router. Some combination of DNS and ARP will
- tell them which to do. Routers _must_ exist of course, e.g. as gateways
- to the non-ATM world, but should be bypassed for efficiency whenever
- possible. There are specific proposals from the IP-over-large-public-
- data-networks working group for shortcut routing.
-
- Sorry, as usual CCITT and their friends are too far behind reality.
- IETF and the ATM Forum will have resolved all these questions before
- CCITT even formally votes on AAL5.
-
- Regards,
- Brian Carpenter CERN, brian@dxcern.cern.ch
- voice +41 22 767 4967, fax +41 22 767 7155
-