choice of file protocol (AFP vs CIFS vs NFS)

As a point of comparison to the prior post, here are the same tests performed using the same client, this time against an older Windows machine as the server. I wanted to see how Microsoft's SMB server compares, still using OS X as the client.


Testbed:

Client testbed: Mac Pro, OS X 10.4.9, 5GB RAM.
Server testbed: homebuilt PC, Athlon XP processor, Windows XP, 1GB RAM.
Network: gigabit ethernet, linksys switch, normal Ethernet MTU (1500).


Test methodology:

Same as above. Testing CIFS only since that's all Windows supports out of the box.


Test results:

Large writes: 1GB, written 1MB at a time.
(dd bs=1048576 count=1024 if=/dev/zero of=/t/afp/testfile)

CIFS: 49.7 sec (21.6 MB/sec) - ReadyNAS was 157.7 sec (6.8 MB/sec)


Small writes: 10MB, written 1KB at a time.
(dd bs=1024 count=10000 if=/dev/zero of=/t/afp/testfile)

CIFS: 3.4 sec (3.0 MB/sec) - ReadyNAS was 7.5 sec (1.4 MB/sec)

Small reads: 100MB, read 1KB at a time.
(dd bs=1024 count=100000 of=/dev/null if=/t/afp/testfile)

CIFS: 36.9 sec (2.8 MB/sec) - ReadyNAS was 64.0 sec (1.6 MB/sec).


Large reads: 1GB, read 1MB at a time.
(dd bs=1048576 count=1024 of=/dev/null if=/t/afp/testfile)

CIFS: 73.6 sec (14.6 MB/sec) - ReadyNAS was 78.6 sec (13.6 MB/sec).


Conclusion:

Again, OS X's CIFS performance is much worse for small reads/writes than large reads/writes. (I did notice one difference: the read tests were faster if performed back to back, for data sizes that are smaller than the server's memory, and not for data sizes that are larger than the client's memory. This might indicate that the server is disk-bound and caching on the server helps. It doesn't seem to indicate client-side caching, because that wouldn't depend on the amount of memory available on the server.)