 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
WARNING: This information has not been updated since October, 1997!
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
INDEX ENTRY FOR NFSSTONES:
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Name: NFSSTONEs - Simple NFS benchmarking utility
Version: not stated
Author(s): Barry Shein <bzs@world.std.com>
Mike Callahan
Paul Woodbury <pww@world.std.com>
Ftp source: ftp.std.com:/src/bench
Description:
This benchmark should be as useful in comparing different
networked file systems (e.g., NFS vs. RFS[4]) as it is in
comparing different implementations, both hardware and software,
of the same networked file system.
We have called the units reported by the benchmark NFSSTONEs.
This is the total number of operations in one run through the
program (45,522 with the default parameters used throughout this
paper) divided by the total elapsed time in seconds. This can
also be thought of as NFS operations per second, where NFS
operations represent a mixture of requests tuned to reflect what
we believe is normal usage (although compressed into a small time.)
Since the mix of operations can be critical the benchmark program
was designed to allow the person performing the tests to easily
vary the proportions of reads, writes, look- ups etc. The default
values, however, were based on a par- ticular mixture obtained by
empirical measurement by [SAND- BERG85].
We consider this choice of default parameters to be suffi-
ciently compelling to recommend them strongly when comparing file
servers. We are happy that, should different parameters be
desired, it is easy to vary the benchmark's values to fit new
models.
-- Adapted from NFSSTONE: A Network File Server Performance
Benchmark, Barry Shein, Software Tool & Die; Mike Callahan, Paul
Woodbury, Encore Computer Corporation, p. 269, USENIX Conference
Proceedings, Summer 1989.
Advertised architectures:
Not stated
Prerequisites:
C compiler, file server
