xfs
[Top] [All Lists]

Re: xfslogd on SMP systems

To: "HABBINGA,ERIK ""(HP-Loveland,ex1)" <erik.habbinga@xxxxxx>
Subject: Re: xfslogd on SMP systems
From: Steve Lord <lord@xxxxxxx>
Date: 03 Apr 2003 16:32:03 -0600
Cc: "'linux-xfs@xxxxxxxxxxx'" <linux-xfs@xxxxxxxxxxx>
In-reply-to: <F341E03C8ED6D311805E00902761278C0C35E4BD@xfc04.fc.hp.com>
Organization:
References: <F341E03C8ED6D311805E00902761278C0C35E4BD@xfc04.fc.hp.com>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Thu, 2003-04-03 at 16:05, HABBINGA,ERIK (HP-Loveland,ex1) wrote:
> I'm running SPEC SFS NFS testing on a 4 way Pentium IV Xeon server.  There
> are 4 xfslogd processes running: xfslogd/0, xfslogd/1, xfslogd/2, and
> xfslogd/3.  I only ever see xfslogd/0 consume any CPU time, and the other
> xfslogd processes do nothing.  Is this the expected behaviour?  Should all 4
> xfslogd processes be sharing the work instead?  Why is there an xfslogd
> process per cpu?
> 
> Thanks,
> Erik Habbinga
> Hewlett Packard

It is being reworked. The threads handle I/O completion on a per cpu
basis, interrupts hand work off to the thread on the same cpu. Xeons
deliver all their interrupts to cpu 0 by default, so all the work
goes one place. Ask intel about this one.

If you actually have multiple controllers involved then you can
get hold of irqbalance which will distribute the interrupts from
different controllers around the cpus.

This xeon issue is why we decided to rework this.

Steve

-- 

Steve Lord                                      voice: +1-651-683-3511
Principal Engineer, Filesystem Software         email: lord@xxxxxxx


<Prev in Thread] Current Thread [Next in Thread>