xfs
[Top] [All Lists]

Re: [review] Remove the xfs refcache

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [review] Remove the xfs refcache
From: Lachlan McIlroy <lachlan@xxxxxxx>
Date: Tue, 18 Dec 2007 15:12:02 +1100
Cc: Donald Douwsma <donaldd@xxxxxxx>, xfs-oss <xfs@xxxxxxxxxxx>, gnb@xxxxxxx
In-reply-to: <20071217071426.GA11462@xxxxxxxxxxxxx>
References: <4765EC66.5020202@xxxxxxx> <4765F444.8010705@xxxxxxx> <20071217071426.GA11462@xxxxxxxxxxxxx>
Reply-to: lachlan@xxxxxxx
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Thunderbird 2.0.0.9 (X11/20071031)
Christoph Hellwig wrote:
On Mon, Dec 17, 2007 at 03:00:04PM +1100, Lachlan McIlroy wrote:
Hey hold on there buddy!  We may need to reactivate this code to fix some
performance issues.  I've acttually got this code working and proven that
it is one way to fix some of the NAS/NFS issues we have.

The little comment that reads "reference cache not needed for NFS in 2.6"
is wrong - we do need it or something like it.

Not in XFS, though.  This thing needs to be done genericly in NFSD.
Greg has been working on an open files cache in nfsd which should be
helping this.

Adding a cache like this back into XFS will get my veto (at least for
mainline where I have a bit of a say :))


Greg's NFS OFC will provide better performance for strictly NFS workloads
and if all we are trying to fix here is the NFS issues then I agree with
you that the OFC should not go into XFS.

Since I have been able to reproduce some of our NAS/NFS performance problems
without NFS (that is demonstrate that the problems are in XFS) it makes some
sense to fix these in XFS.  I have observed that for some non-NFS workloads
we see a significant reduction in log traffic with the OFC in XFS so for
reasons beyond NFS there may be a need to reactivate the refcache code.  For
the moment we are still analysing the pros/cons.


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