Academic Workstations

[Kemp at DOCKMASTER.NCSC.MIL]

Barry Shein writes:
 > However, I will agree that blaming it on the diskless workstations is
 > a wonderful alibi, the yokels believe you and rarely ask you to
 > actually do your job and find out what's really causing the problem.
 >
 > It's the diskless workstations (we know the diskless workstation users
 > will never buy the local disks you recommend so it's a safe bet to blame
 > it on them).

Hey Barry,
 I guess I don't understand the academic environment, where apparently
someone (a computer center?)  buys the server and someone else (a
yokel?)  buys the clients.  In my environment, where total dollars (your
tax dollars!)  spent is the bottom line, it is both economically and
technically advantageous to use dataless clients.

 A few years ago, it was OK for us to have a 3/160 server with a fast
SMD disk (I use the term "fast" *very* loosely to refer to the Xylogics
451) serving three diskless 3/110 clients, but that strategy doesn't
scale well.  We now have a eleven diskless clients running off of two
servers, and that scheme is beginning to creak.  Our latest buy is for a
bunch of Sparcstations and enough 600 MB SCSI's to convert all the
clients to dataless nodes.  At $3150 each, the 16 ms, 4 MB/s local disks
seemed like a bargain compared to the bucketfull of money we could have
paid for the 4/390 fileserver with IPI disks.

 Your point about CPU parallelism and namei is valid, but in general it
seems self evident that the parallelism of multiple drives (SCSI busses
& head seeks) is a much more important factor favoring the distributed
approach.  It is easier to stock spares for cheap client disks than for
expensive server disks.  Also, our servers occasionally crash or are
rebooted or are used for tape operations, and having clients hang
waiting for swapping is very disruptive.

  Users home directories are on the server, but private files can be
kept on the local disk as long as the user understands that he is
responsible for backing it up.  That will eliminate grumbling over who
is hogging disk space on the server, always a problem in our shop.

  As to the other technical issues, Brent Byer seems to have summed up
my feelings pretty well.

  Dave Kemp