[comp.dcom.lans] Mitigating against server failure.

romeo@lindy.Stanford.EDU (Patrick Goebel) (12/07/89)

We are a small institution on the verge of installing a local area
network with a Sun server at the hub, a couple of Sun workstations,
about 45 IBM PC's and approximately 25 Mac's.

In particular, we have in mind a SPARCserver 370 (32Mb RAM, 688+ Mb
SMD hard disk capacity) for file, mail and printer service and two
SPARCstation 1's for dedicated statistical analysis and numerical
simulations.  For tape backup we are considering one of the 8mm
systems such as Delta Microsystem's ExoByte tape drive.

Our server will connect to the Internet via SUnet through a Cisco box
and a 56K baud DDS line.  Our PC's and Sun's will communicate over
ethernet, while the Mac's will have a choice between ethernet and
Appleshare.

I would be extremely grateful for suggestions concerning the following
points:

(1) What is the best way to recover from a server failure?  We will
have a hardware maintenance contract with Sun but would this be the
fastest way to get the system back on at least one leg?  For example,
in the event of a disk failure, would it be possible to dump a backup
tape onto one of the SPARCstation 1 disks and use the workstation as a
temporary server?

(2)  Are we being paranoid, or are server failures fairly common?

(3) Are there special problems related to running workstations with
their own disks as opposed to diskless?  (It was once suggested to me
that diskFULL clients can sometimes be a source of confusion in a
network environment.)

(4) If the server goes down, what effect, if any, does this have on
the ability of a PC user to run a terminal emulation and connect to
the outside world through the Cisco box?  In other words, is the 56K
baud connection to the outside world logically independent of the
server?

Many thanks in advance,
patrick goebel
budding system administrator
CASBS