I'm using a dual P-III Compaq DL380 running Red Hat 8.0 to host a new
ATABoy RAID array purchased from Nexsan. It's connected to our switch
via a gig/fiber card based on the Intel 82542 chipset.
The service stack (for lack of a better phrase) is
* ATABoy2 -- six-drive RAID-5 array that presents itself to the host
as a single SCSI device (no multiple LUNs)
* LVM -- 1.0.3, one PV, one VG, one LV (so far :-)
* XFS -- 1.2pre5
* NFS -- v2/v3, UDP only, 80 nfsd threads
* Samba -- 2.2.7
Last night, I set up six NFS clients and one Linux/smbfs client to run
iozone against the nfs mount/samba share.
I was keeping track of stats using iostat -x on the server. My typical
wkB/s numbers hovered in the 15000 to 19000 range. Is that sort of
result set consistent with what any of you have seen for Linux-hosted
NFS RAID exports?
--Paul Heinlein <[email protected]>
-------------------------------------------------------
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 - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
hi paul-
you can check the list archives for more information,
but the experiences on this list have been that RAID
5 is generally not a hi-performance back-end solution
for Linux NFS servers.
i believe the problem is that NFS servers generate lots
of small writes, and that's something RAID 5 is not
very good at.
i can't answer your specific question about the numbers
you posted. maybe someone else?
> -----Original Message-----
> From: Paul Heinlein [mailto:[email protected]]=20
> Sent: Friday, February 14, 2003 10:32 AM
> To: [email protected]
> Subject: [NFS] iostat wkB/s numbers during testing
>=20
>=20
> I'm using a dual P-III Compaq DL380 running Red Hat 8.0 to=20
> host a new ATABoy RAID array purchased from Nexsan. It's=20
> connected to our switch via a gig/fiber card based on the=20
> Intel 82542 chipset.
>=20
> The service stack (for lack of a better phrase) is
>=20
> * ATABoy2 -- six-drive RAID-5 array that presents itself to the host=20
> as a single SCSI device (no multiple LUNs)
> * LVM -- 1.0.3, one PV, one VG, one LV (so far :-)
> * XFS -- 1.2pre5
> * NFS -- v2/v3, UDP only, 80 nfsd threads
> * Samba -- 2.2.7
>=20
> Last night, I set up six NFS clients and one Linux/smbfs=20
> client to run=20
> iozone against the nfs mount/samba share.
>=20
> I was keeping track of stats using iostat -x on the server.=20
> My typical wkB/s numbers hovered in the 15000 to 19000 range.=20
> Is that sort of result set consistent with what any of you=20
> have seen for Linux-hosted=20
> NFS RAID exports?
>=20
> --Paul Heinlein <[email protected]>
>=20
>=20
>=20
> -------------------------------------------------------
> This SF.NET email is sponsored by: FREE SSL Guide from=20
> Thawte are you planning your Web Server Security? Click here=20
> to get a FREE Thawte SSL guide and find the answers to all=20
> your SSL security issues.=20
> http://ads.sourceforge.net/cgi-> bin/redirect.pl?thaw0026en
>=20
>=20
> _______________________________________________
> NFS maillist - [email protected]=20
> https://lists.sourceforge.net/lists/listinfo/n> fs
>=20
-------------------------------------------------------
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 - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
If you do lots of writes, it's especially important
that you enable the RAID controllers write-back cache
(if you're concerned about data loss in case of power
failure, you can get battery backup for most cards for
a little more than $100). Also, ideally you'd try to
keep write-intensive stuff on a RAID-1 partition, and
the rest on RAID-5, b/c RAID-5 is _very_ inefficient
at writes. If you have 6 drives, why don't you set 2
of them aside for RAID-1 and the most write-intensive
files?
--- "Lever, Charles" <[email protected]> wrote:
> hi paul-
>
> you can check the list archives for more
> information,
> but the experiences on this list have been that RAID
> 5 is generally not a hi-performance back-end
> solution
> for Linux NFS servers.
>
> i believe the problem is that NFS servers generate
> lots
> of small writes, and that's something RAID 5 is not
> very good at.
>
> i can't answer your specific question about the
> numbers
> you posted. maybe someone else?
>
> > -----Original Message-----
> > From: Paul Heinlein [mailto:[email protected]]
> > Sent: Friday, February 14, 2003 10:32 AM
> > To: [email protected]
> > Subject: [NFS] iostat wkB/s numbers during testing
> >
> >
> > I'm using a dual P-III Compaq DL380 running Red
> Hat 8.0 to
> > host a new ATABoy RAID array purchased from
> Nexsan. It's
> > connected to our switch via a gig/fiber card based
> on the
> > Intel 82542 chipset.
> >
> > The service stack (for lack of a better phrase) is
> >
> > * ATABoy2 -- six-drive RAID-5 array that presents
> itself to the host
> > as a single SCSI device (no multiple
> LUNs)
> > * LVM -- 1.0.3, one PV, one VG, one LV (so far
> :-)
> > * XFS -- 1.2pre5
> > * NFS -- v2/v3, UDP only, 80 nfsd threads
> > * Samba -- 2.2.7
> >
> > Last night, I set up six NFS clients and one
> Linux/smbfs
> > client to run
> > iozone against the nfs mount/samba share.
> >
> > I was keeping track of stats using iostat -x on
> the server.
> > My typical wkB/s numbers hovered in the 15000 to
> 19000 range.
> > Is that sort of result set consistent with what
> any of you
> > have seen for Linux-hosted
> > NFS RAID exports?
> >
> > --Paul Heinlein <[email protected]>
> >
> >
> >
> >
>
-------------------------------------------------------
> > 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 - [email protected]
> > https://lists.sourceforge.net/lists/listinfo/n> fs
> >
>
>
>
-------------------------------------------------------
> 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 - [email protected]
> https://lists.sourceforge.net/lists/listinfo/nfs
__________________________________________________
Do you Yahoo!?
Yahoo! Shopping - Send Flowers for Valentine's Day
http://shopping.yahoo.com
-------------------------------------------------------
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 - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
On Fri, 14 Feb 2003, Alan Powell wrote:
> If you do lots of writes, it's especially important that you enable
> the RAID controllers write-back cache
The ATABoy ships with a 128MB write cache, enabled by default.
> Also, ideally you'd try to keep write-intensive stuff on a RAID-1
> partition, and the rest on RAID-5, b/c RAID-5 is _very_ inefficient
> at writes. If you have 6 drives, why don't you set 2 of them aside
> for RAID-1 and the most write-intensive files?
The filesystem we're exporting is for general-purpose use and is
directory-divided among a number of customer groups. If I were
setting the export up for a specific application, then your suggestion
would make a whole lotta sense. I'll certainly keep it in mind if one
or more customers have specific complaints about write performance.
--Paul Heinlein <[email protected]>
-------------------------------------------------------
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 - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs