Jamison Collins (indecent_and_obscene@yahoo.com)
Thu, 4 May 2000 00:48:54 -0700 (PDT)

bleh, only did Sprite...

The Sprite Network OS was implemented in order to
attempt to take advantage of 3 computer trends—the
increasing commonality of multiple workstations
connected on a local network (as opposed to a single
time sharing system), the increasing memory in
workstations and the emergence of multiprocessor
workstations. This is an important work because it
attempts to reevaluate the design of an OS when these
new trends are taken into account.
Specific decisions were made to take advantage of
each trend. In order to take advantage of the local
network of workstations, the primary design decision
was to implement process migration and a well
implemented ‘third generation’ network transparent
filesystem. To take advantage of future
multiprocessing capabilities the kernel was written in
a different structure such that a single kernel call
would not lock out any further kernel calls. Finally,
to take advantage of the increase in available memory,
both workstations and file servers had large file
caches. The authors even make the comment that the
file cache at the servers enables files to be served
across the network more quickly than could be read off
a local disk.
Finally, we’re getting into a realm to which I am
accustomed—performance numbers. They’re not IPC, but
I guess they’ll have to do. We see that a fairly high
bandwidth connection is achievable between a
workstation and a file server. Additionally, through
judicious use of process migration, the time to
compile a program is greatly reduced. Finally,
through the caching of files, the network is able to
sustain a much higher workload than existing
approaches. Thus, through experimentation and
benchmarking, we see that Sprite does indeed fulfill
most of its promises.
This was a pretty interesting paper. I was able to
follow it well because it had a ‘modern’ feel more to
it than other papers. I am, however, unimpressed with
some of the reasoning of the authors. As I recall, in
1992, my pc had at least an order of magnitude larger
hard drive than I had ram. I understand that the
workstation market differs from the pc market, but I
don’t understand why the authors felt that available
memory would grow so quickly, and even to the point of
optimizing the OS for the case of workstations even
without local storage.


Do You Yahoo!?
Send instant messages & get email alerts with Yahoo! Messenger.