From: "Heflin, Roger A." <Roger.A.Heflin@conocophillips.com>
Subject: Re: nfsd tuning - please help me! (Alan Powell)
Date: Fri, 14 Feb 2003 15:10:28 -0600
Sender: nfs-admin@lists.sourceforge.net
Message-ID: <5CA6F03EF05E0046AC5594562398B916A3281B@POEXMB3.conoco.net>
Mime-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Cc: <lakerfaniam2@yahoo.com>
To: <nfs@lists.sourceforge.net>
Errors-To: nfs-admin@lists.sourceforge.net


>  Date: Fri, 14 Feb 2003 09:44:53 -0800 (PST)
> From: Alan Powell <>
> Subject: Re: [NFS] nfsd tuning - please help me!
> To: Steve Dickson <SteveD@RedHat.com>, nfs@lists.sourceforge.net
>=20
> Unfortunately, we've tried all that already. So given
> that we are not hardware/network constrained, does all
> this mean that the Linux kernel NFS runs into
> performance issues beyond 100 file reads/sec?
>=20
>=20
	I have been able to get closer to 10-20MBytes per second with
	linux nfs.  The netapps will do around 4-5 times that though at
	a higher cost.  And you can get it out of linux, by putting more
	cheap smaller servers to obtain the same rate.

	What are you underlying disks?  You could still be hardware
	constrained depending on what your underlying disks are,
	and what you underlying disk controller is.   Both can have
	issues.

	I have machines that are servicing around 2500 8k reads
	per second and seem to work fine, though mine may break
	down to fewer larger reads.

	Other things that will get you in trouble is having lots of files
	in a single directory (in the several thousand range will hurt
	pretty bad), also check to make sure you aren't accumulating lots
	of .nfs* files in the directies in question, I had a situation where
	there where lots of files being messed with (read and write) and
	lots of these files accumulated and pretty much brought
	performance to its knees.  The solution was to run a cron job
	to clean up the .nfs* files.  The .nfs files are created when you
	are reading a file that is being deleted by another process at
	the same time, the .nfs* stays around to service the reader,
	and does not always go away (this is on all NFS implementations
	I have seen).

	Do a ls -ld dirname and see the size of the directories, and include it
	in the next message if one of the above don't pan out.

							Roger

				=09


-------------------------------------------------------
This SF.NET email is sponsored by: FREE  SSL Guide from Thawte
are you planning your Web Server Security? Click here to get a FREE
Thawte SSL guide and find the answers to all your  SSL security issues.
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0026en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs