Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!haven!mimsy!chris
From: chris@mimsy.UUCP (Chris Torek)
Newsgroups: comp.dcom.lans
Subject: Re: broadband high speed networking between buildings
Message-ID: <19691@mimsy.UUCP>
Date: 20 Sep 89 13:19:06 GMT
References: <4616@ursa-major.SPDCC.COM>
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 33

>In article <19661@mimsy.UUCP> I wrote:
>>broadband is slow.

In article <4616@ursa-major.SPDCC.COM> eli@spdcc.COM (Steve Elias) writes:
>wrong.  10 Mbits/sec is not "slow", by today's standards.

By *whose* standards?

>... it is rare to see a network which pushes the limits of
>a 10 Mbits/sec ethernet, you know!

Rare for you, perhaps.

10 Mb/s is only 1.25 MB/s.  That is slow.  If you have 100 hosts
sharing the network, and each is 1% busy, your available bandwidth is
at best 1.25 MB/s.  (Since network traffic tends to be bursty, it
actually varies.)  In any case, old workstations (Sun 3s) with current
software (Jacobson's header-prediction tcp) already use 1 MB/s (8
Mbits/s) on Ethernet.  Tomorrow's workstations, which (with the
exception of a few places) will be running yesterday's software,
should be able to do the same.  Once you put today's software on
them. . . .

Look at it another way: today's small computer disk drives will
move more than 3 MB/s (on synchronous SCSI).  Today's file systems
cut this in half (alas!).  But this means your file server should
be able to `want' 1.5 MB/s when talking to a single host.  10 Mb/s
is thus a recipe for bottlenecks: today, when fusing two LANs into
one, and tomorrow, as a LAN itself.  (Good thing FDDI still looks
promising.)
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163)
Domain:	chris@mimsy.umd.edu	Path:	uunet!mimsy!chris