Re: [Evolution] Slow filtering under NFS



On Thu, 2002-05-30 at 21:51, Jonathan F. Dill wrote:
What Linux kernels are involved?  Are all of the Linux updates applied? 
Have you tried fs benchmark like "bonnie" to see how fs access to the
NFS server is in general independent of evolution?

I'm running 2.4.9 on all of the machines involved.  Here is the output
from bonnie (miller is the server, disston is the client)

[tad miller ~]$ bonnie -s 1500
Bonnie 1.4: File './Bonnie.28605', size: 1572864000, volumes: 1
Writing with putc()...         done:  15068 kB/s  85.0 %CPU
Rewriting...                   done:  16211 kB/s  18.9 %CPU
Writing intelligently...       done:  25741 kB/s  25.5 %CPU
Reading with getc()...         done:  13346 kB/s  77.6 %CPU
Reading intelligently...       done:  54586 kB/s  35.2 %CPU

[tad disston ~]$ bonnie -s 1500
Bonnie 1.4: File './Bonnie.1755', size: 1572864000, volumes: 1
Writing with putc()...         done:   4338 kB/s  52.5 %CPU
Rewriting...                   done:   2265 kB/s  11.4 %CPU
Writing intelligently...       done:   4543 kB/s  12.9 %CPU
Reading with getc()...         done:   4894 kB/s  54.5 %CPU
Reading intelligently...       done:   4820 kB/s   7.4 %CPU


Clearly NFS is much slower than native, but it doesn't feel like the
real source of the slowness (filtering on miller is nearly
instantaneous, filtering on disston is 100s or 1000s of times slower). 
I have tried NotZed's suggestion about turning off indexing and will
benchmark it once I have accumulated enough mail to make it a worthy
test.



to do with file locking and/or blocking I/O.  Why not just use plain
IMAP to get to the stuff on the server?  File locking with e-mail is a

Two main reasons.

1) NFS is already up and running.

2) The last 7 errata from Redhat have been related to security issues
with IMAP.  I like email to be one of those things that just works and
doesn't require lots of monitoring of the security patch of the week.


In the end, I'll probably end up with IMAP, but for the short term I'm
going to continue poking at the current setup.


Thanks for the info,
Tad





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]