lkml.org 
[lkml]   [1999]   [Mar]   [15]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: Poor read NFS performance
From
Date
danis@sagem.fr (Thierry Danis) writes:

> Hello,
>
> I run iozone between two plain 2.2.3 boxes (server and client) :
>
> /willy/u1/tmp > ~/iozone/iozone-2.2.1-ac3 200 8192 zz
>
> IOZONE: Performance Test of Sequential File I/O -- V2.01 (10/21/94)
> By Bill Norcott
>
> Operating System: POSIX 1003.1-1990
>
> Send comments to: b_norcott@xway.com
>
> IOZONE writes a 200 Megabyte sequential file consisting of
> 25600 records which are each 8192 bytes in length.
> It then reads the file. It prints the bytes-per-second
> rate at which the computer can read and write files.
>
>
> Writing the 200 Megabyte file, 'zz'...63.280000 seconds
> Reading the file...242.180000 seconds
>
> IOZONE performance measurements:
> 3314083 bytes/second for writing the file
> 865947 bytes/second for reading the file
> /willy/u1/tmp >
>
> The machines are hooked with a 100 Mb/s ethernet card.
> The automount options specify : rsize=8192,wsize=8192.
>
> With 2.2.1-ac7, I had around 3,5 Mb/s during reading. 860 Kb/s is
> really bad.
>
> Is that related to the 4k page size discussed in the list ?
>

Nope. The changes to the NFS client between 2.2.1-ac7 and 2.2.3-ac1
are mainly in the form of RPC flow control.

I saw your previous posting about 'new' error messages of the form
'server xxx not responding'. Is this still the problem that is slowing
things down? The reason for those messages was that the NFS client is
being told that the server is not listening (error =
-ECONNREFUSED). There are 3 possible causes: either a bug in the Linux
UDP layer, or a bug in your server's UDP layer, or you are using too
few NFSd daemons. In principle ECONNREFUSED is only supposed to be
returned when the server is rebooting or something like that.

If you could produce a tcpdump of what is going on, that would be very
helpful...

Amicalement,
Trond

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:50    [W:0.042 / U:1.824 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site