[595] in Info-AFS_Redistribution
NFS vs. AFS builds
daemon@ATHENA.MIT.EDU (dstokes@sunlight.austin.ibm.com)
Tue Feb 11 16:27:57 1992
Date: Tue, 11 Feb 92 14:04:25 -0600
From: dstokes@sunlight.austin.ibm.com
To: info-afs@transarc.com
Cc: carl@shade.austin.ibm.com
Nobody here in Austin has come up with a good answer for these questions,
so I thought I post this to the experts. Any thoughts???
Thanks, Dawn
From awdprime!anderson.austin.ibm.com!craig Tue Feb 11 13:59:41 CST 1992
Article: 514 of ibm.awd.afs
Path: awdprime!anderson.austin.ibm.com!craig
From: craig@anderson.austin.ibm.com (Craig Anderson)
Newsgroups: ibm.awd.afs
Subject: AFS 2-8 times slower than NFS for builds.
Message-ID: <16133@awdprime.UUCP>
Date: 11 Feb 92 18:26:51 GMT
Sender: news@awdprime.UUCP
Reply-To: craig@aixwiz.austin.ibm.com
Distribution: ibm
Organization: IBM AWD, Austin
Lines: 42
AFS is slower to do builds. A lot slower. Why is this? Why doesn't
a 200MB cache fix this? Why is the 2nd build no faster than the first?
The following is a comparison of AFS vs. NFS builds*. The net result
here is that AFS took twice as long as NFS. More importantly, this
example is compile bound (for NFS anyway)...
Real User Sys Comments
AFS 1st time 7:54.13 2:42.61 3:49.02 ~82% CPU utilizaton
2nd time ** 7:54.49 2:42.77 3:48.24 ~82% CPU utilizaton
NFS 1st time 4:15.01 2:15.38 25.71 ~64% CPU utilizaton
2nd time 4:04.51 2:15.29 26.31 ~66% CPU utilizaton
Below is a comparison on non-compile build performance (target is up to date).
Time for lm to determine "Target rsdd is up to date.". This is similar
to the time from execution of "lm" command to start of first compile
(AFS is much much slower here). AFS took about 4.5 longer than NFS
(and consumed about 8 times more CPU time):
Real User Sys Comments
AFS ** 3:54.14 18.78 3:13.05 ~91% CPU utilizaton
AFS ** 3:44.72 18.99 3:12.77 ~95% CPU utilizaton
NFS 49.40 10.85 14.69 ~51% CPU utilizaton
NFS 46.73 10.73 14.94 ~55% CPU utilizaton
* Both examples use "lm" to build the rsdd device driver using
"timex lm FORCE=FORCE list=list rsdd". The benchmarks above on a 520 with
24MB Ram, and a 200MB AFS cache on a separate drive. Little else was running.
The source for both the AFS and NFS code and build environments was located
on the same machine (speedo). Ping typically takes 9ms to this machine.
** For these cases one would expect it to be faster since AFS should
be using the cache to improve performance.
-Craig
--
Craig Anderson (craig@aixwiz.austin.ibm.com) JIS:$B%/%l%$%0 %"%s%@!<%=%s(B
These are my personal views and do not reflect policy or commitments of IBM.
