In the same spirit, I ran some comparative tests of nfs client vs smbfs
vs ext2, which produced the results below.
In the tests, smbfs was talking to an NT 4.0 server (P6 200, 64M) over
100mb ethernet with no traffic, while nfs was talking to the unfsd
server on the same machine via loopback.
-------Sequential Output-------- ---Sequential Input--
--Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block---
--Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU
/sec %CPU
ext2 10 3046 53.9 3514 16.5 3277 36.8 4954 99.7 40302 98.4
1542.1 35.1
smbfs 10 286 33.6 1457 28.8 1721 37.3 6108 95.4 40243 98.3
1845.3 53.1
nfs client 10 1487 30.2 2235 6.5 1113 9.7 1768 32.3 2479 17.7
535.5 19.5
The size of the test file is small enough to be effectively cached on my
system (32M), and the excellent read performance of smbfs shows its
effective use of page cache. The poor per-char writing speed is a
result of strictly synchronous writes -- smbfs has to write a page each
time it gets dirtied.
I'm not sure why the nfs read performance is so low -- will have to look
into this.
Regards,
Bill