slow performance -- how to diagnose
Subsystems (2)
I'm experiencing some slow performance that I have done some initial diagnosis as per http://nfs.sourceforge.net/nfs-howto/performance.html
but not sure where to go from here.
I have tried both UDP and TCP and in fact UDP performance seems to be slightly worse.
The connection between a client and the server is gige and seems to perform pretty close to such using ttcp:
ttcp-t: 819200000 bytes in 7.368 real seconds = 848.299 Mbit/sec +++
ttcp-t: 819200000 bytes in 7.01 real seconds = 891.19 Mbit/sec +++
The first is from client to server and the second server to client. Both seem reasonable.
The speed of the disk on the server using a simple dd test:
# time dd if=/dev/zero of=file bs=1M count=4000 4000+0 records in 4000+0 records out
real 2m28.756s user 0m0.024s sys 0m17.141s
(1024*1024*4000)/148.7/(1024*1024)*8 == 26.899 MB/sec == 215.192Mbit/s
But from an NFS client...
# rsh client5 "(cd $(pwd);time dd if=/dev/zero of=file bs=1M count=1000)" 1000+0 records in 1000+0 records out
real 1m21.036s user 0m0.006s sys 0m1.982s
(1024*1024*1000)/81.036/(1024*1024) == 12.340 MB/sec
Any ideas where to look?
b.
Signature · 3 lines
My other computer is your Microsoft Windows server. Brian J. Murrell