NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/comp/dcom/cellrel/435

Text File | 1992-08-31 | 2.6 KB | 53 lines

Newsgroups: comp.dcom.cell-relay
Path: sparky!uunet!pipex!warwick!str-ccsun!ccsun.strath.ac.uk!craa85
From: [e-mail] ( D.W.Stevenson)
Subject: Re: Future of IP routers
Message-ID: <[e-mail]>
Sender: [e-mail] (News account )
Nntp-Posting-Host: coll
Organization: University of Strathclyde
References: <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]>
Date: Mon, 31 Aug 1992 15:15:29 GMT
Lines: 40

In article <[e-mail]>, [e-mail] (Randall Atkinson) writes:
|>
|>   Not at all the same function.  IP performs dynamic routing with the
|> path traversed (by packets from source A to destination B) changing
|> all the time based on loading, congestion, and whatever.  With ATM,
|> routing is done only once (from source A to destination B) -- only at
|> call setup time -- and so ATM does switching at switches rather than
|> routing at switches.  If the ATM path chosen at call setup time has a problem
|> (e.g. a switch fails) then connectivity is completely lost.  If an IP router
|> dies, the other routers will automatically and dynamically route around it
|> until it comes back up and connectivity is not lost.

I agree with what you say except IP is not responsible for providing
a guaranteed delivery. It is the responsibility of the transport layer to ensure
delivery. Therefore even if an ATM connection dies, the transport layer
will request a retransmission and hence a new connection,
perhaps via a different switch,
and retransmit up to the point where the connection died.

The public ATM network will presumably have the same level of
switch and route redundancy as current telco networks
and so an alternate route should be available if a switch fails. The overhead
in re-establishing a connection will be small and I presume the rate
of switch failure will be small also. These are the sort of things that
only experience will show.

One feature of the ATM network is that it has very low error rates
since it's running over fibre all the
way. Congestion will hopefully not be too much of a problem (but don't
quote me on that!) since bandwidth is requested at call set-up time.

I do understand what you are saying though - routers separate
a huge connection orientated network into a number of smaller 'connection
domains' (my words) which will increase fault tolerance in large internetworks.

--
Dave Stevenson                 [e-mail]
Computer Centre Communications        Tel : [phone] ext 3461
University of Strathclyde
Glasgow, Scotland, U.K.