Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!CIM-VAX.HONEYWELL.COM!derstad
From: derstad@CIM-VAX.HONEYWELL.COM ("DAVE ERSTAD")
Newsgroups: comp.sys.apollo
Subject: Re:  Academic workstations
Message-ID: <8906121253.AA01905@umix.cc.umich.edu>
Date: 12 Jun 89 13:49:00 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 27


The comments about the supiority of a central file server to local
disks are a bit misguided.  Sure, if you have 5000 AT's around,
there's going to be a lot of sneaker-netting.  But, with a real
network, such as Apollo's token ring, you don't care where the 
files are;  you don't need to make local copies;  etc.

Essentially, any reasonable network with a network wide naming
space should give all the advantages of a central file system
without losing the cost and performance advantages of a distributed
file system.

Even if you assume that data is not being used locally (and in 
a large network this is frequently true) you have 50, 100, or 
whatever workstations gatewaying data onto the net from 
however many disks, rather than banging on a couple of
servers.  Plus, your overall net traffic decreases since
you typically have local copies of operating systems and
applications software.

I feel the above will remain true until we have significantly
faster networks (FDDI) and maybe even past then.

Dave Erstad
DERSTAD@cim-vax.honeywell.com
Principal Design Automation Engineer
Honeywell SSEC