nfs performance problem

nfs performance problem

[poczta.dotcom.pl]

[-- Attachment #1: Type: text/plain, Size: 1410 bytes --]

Hi All

I have really slow working nfs.
Server is build on:
- dual pentium Xeon 2GHz,
- 2 3ware 7500 controllers,
- raid 5 on 7 160GB discs on controler 1 
- raid 5 on 6 160GB disks on cotroller 2 

Serwer is used for ~500k maildirs, 
all clients (~20) are freebsd boxes, when nfs transfer is around 1MB/s for writing and 1.5MB/s for reading 
listing user directories even with very few file stakes up to 15 seconds.

I tried nfs-ALL patch, but it didn' t help.

I run 256 nfs daemons,

echo 2097152 > /proc/sys/net/core/rmem_max
echo 2097152 > /proc/sys/net/core/rmem_default


nsfstat output:

Server rpc stats:
calls      badcalls   badauth    badclnt    xdrcall
25271043   17885061   0          17885061   0

Server nfs v3:
null       getattr    setattr    lookup     access     readlink
25      0% 1159    0% 89718   0% 11702395 46% 10953437 43% 34868   0%
read       write      create     mkdir      symlink    mknod
925727  3% 226250  0% 44964   0% 1089    0% 2299    0% 0       0%
remove     rmdir      rename     link       readdir    readdirplus
35040   0% 42      0% 121064  0% 535     0% 534720  2% 0       0%
fsstat     fsinfo     pathconf   commit
585179  2% 36      0% 0       0% 12493   0%

mount over tcp gives frequent errors on freebsd clients side:
kernel: nfs send error 32 from nfs server xxx.xxx.xxx.xxx:/export

thanks for any help, 

rafal


[-- Attachment #2: Type: text/html, Size: 3748 bytes --]

Re: nfs performance problem

[Ragnar Kjørstad]

On Tue, Nov 05, 2002 at 07:03:43PM +0100, poczta.dotcom.pl wrote:
> Hi All
>=20
> I have really slow working nfs.
> Server is build on:
> - dual pentium Xeon 2GHz,
> - 2 3ware 7500 controllers,
> - raid 5 on 7 160GB discs on controler 1=20
> - raid 5 on 6 160GB disks on cotroller 2=20
>=20
> Serwer is used for ~500k maildirs,=20
> all clients (~20) are freebsd boxes, when nfs transfer is around 1MB/s =
for writing and 1.5MB/s for reading=20
> listing user directories even with very few file stakes up to 15 second=
s.
>=20
> I tried nfs-ALL patch, but it didn' t help.

How is local performance?
What filesystem do you use?
What's the output from "iostat -x -d -k 60 3"?



--=20
Ragnar Kj=F8rstad
Big Storage


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[poczta.dotcom.pl]

----- Original Message -----
From: "Ragnar Kj=F8rstad" <nfs@ragnark.vestdata.no>
To: "poczta.dotcom.pl" <myciel@dotcom.pl>
Cc: <nfs@lists.sourceforge.net>
Sent: Tuesday, November 05, 2002 8:17 PM
Subject: Re: [NFS] nfs performance problem


> On Tue, Nov 05, 2002 at 07:03:43PM +0100, poczta.dotcom.pl wrote:
> > Hi All
> >
> > I have really slow working nfs.
> > Server is build on:
> > - dual pentium Xeon 2GHz,
> > - 2 3ware 7500 controllers,
> > - raid 5 on 7 160GB discs on controler 1
> > - raid 5 on 6 160GB disks on cotroller 2
> >
> > Serwer is used for ~500k maildirs,
> > all clients (~20) are freebsd boxes, when nfs transfer is around 1MB/=
s
for writing and 1.5MB/s for reading
> > listing user directories even with very few file stakes up to 15
seconds.
> >
> > I tried nfs-ALL patch, but it didn' t help.
>
> How is local performance?

local read/writes are ok

> What filesystem do you use?

reiserfs on top of  lvm (to be able to get snapshots)

> What's the output from "iostat -x -d -k 60 3"?
>
below is output from 'iostat -d 60 3'  (- my iostat does not take '-k'
option,
and '-d' gives me empty output) - mayby this is enough?

Linux 2.4.19 (hostname)        11/05/02

Device:        tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
dev8-0       16.16       205.52        82.84    3481842    1403528
dev8-1      160.58      1266.39      1346.12   21455002   22805744

Device:        tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
dev8-0        9.23       134.93        16.13       8096        968
dev8-1      150.92       753.20      1715.87      45192     102952

Device:        tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
dev8-0       10.02       135.07        26.53       8104       1592
dev8-1      152.78       763.33      1767.20      45800     106032


rafal



-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Matt Heaton]

I have 3ware cards as well (2 7500 Series 8 drive), just like yours.  ALso
with 120 GIG drives (7200)
RPM.  Whever I get 150-200 tps in iostat then my NFS runs SO SLOW.  I also
only get 1 MB, to 1.5 MB
over NFS.  The local speed seems to be ok, not great though.   MY OPINION is
that this is because of seek time on the raid
array.   I am serving very small files, just like you are.  I am requesting
about 200-300 files per second from
each NFS server.  So even though our throughput of only 1.5 MB isn't high.
The number of files per second is
actually quite high, and causes things to slow down because of seek time
issues.  PLEASE GIVE US CACHEFS SOMEONE??

Does anyone have experience with IDE Raid arrays that get over 250 tps in
iostat that work fine?  I would
be VERY VERY VERY interested to find out.

L8r...

Matt


> On Tue, Nov 05, 2002 at 07:03:43PM +0100, poczta.dotcom.pl wrote:
> > Hi All
> >
> > I have really slow working nfs.
> > Server is build on:
> > - dual pentium Xeon 2GHz,
> > - 2 3ware 7500 controllers,
> > - raid 5 on 7 160GB discs on controler 1
> > - raid 5 on 6 160GB disks on cotroller 2
> >
> > Serwer is used for ~500k maildirs,
> > all clients (~20) are freebsd boxes, when nfs transfer is around 1MB/s
for writing and 1.5MB/s for reading
> > listing user directories even with very few file stakes up to 15
seconds.
> >
> > I tried nfs-ALL patch, but it didn' t help.
>
> How is local performance?

local read/writes are ok

> What filesystem do you use?

reiserfs on top of  lvm (to be able to get snapshots)

> What's the output from "iostat -x -d -k 60 3"?
>
below is output from 'iostat -d 60 3'  (- my iostat does not take '-k'
option,
and '-d' gives me empty output) - mayby this is enough?

Linux 2.4.19 (hostname)        11/05/02

Device:        tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
dev8-0       16.16       205.52        82.84    3481842    1403528
dev8-1      160.58      1266.39      1346.12   21455002   22805744

Device:        tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
dev8-0        9.23       134.93        16.13       8096        968
dev8-1      150.92       753.20      1715.87      45192     102952

Device:        tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
dev8-0       10.02       135.07        26.53       8104       1592
dev8-1      152.78       763.33      1767.20      45800     106032


rafal



-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs






-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Benjamin LaHaise]

On Tue, Nov 05, 2002 at 01:22:25PM -0700, Matt Heaton wrote:
> each NFS server.  So even though our throughput of only 1.5 MB isn't high.
> The number of files per second is
> actually quite high, and causes things to slow down because of seek time
> issues.  PLEASE GIVE US CACHEFS SOMEONE??

How is cachefs going to help?  The kernel is already trying to cache data 
as much as possible.  Once you're trying to serve more data than you have 
RAM, this are naturally going to degreate quite significantly as the system 
becomes seek bound.

> Does anyone have experience with IDE Raid arrays that get over 250 tps in
> iostat that work fine?  I would
> be VERY VERY VERY interested to find out.

Use raid1+0 and you'll be much happier, as read requests will be balanced 
over multiple drives (mirroring means the same data can be read from all 
of the mirrors).  Additionally, you'll have much lower CPU utilization 
and writes won't cause all disks in the array to seek for strip updates.  
Read the archives for the past couple of weeks for another example of the 
performance increase when going from raid5 to raid1+0.

		-ben


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Matt Heaton]

Cachefs will help quite a lot in my opinion because it doesn't just store
the files in RAM,
it uses the hard drive.  So if you have an NFS client with an extra 5 gig
that you can
designate as cache then reads to the NFS server will go down DRAMATICALLY as
it will hit local cache on the NFS clients drive.

I agree raid 1+0 should be much faster for writes and a little for read, but
RAID 5 still
reads from all drives simultaneously (Has to read parity in too I know), but
can read
all 7 drives at once instead of only 4 drives at once in a raid 1+0
configuration with 8 drives
in the array.  I have never used 1+0 so I am only talking about physical
drive layout rather
than any personal experience.  Are my assumptions correct that raid 5 does
in fact read
from all drives at the same time?  If so, reading might be a LITTLE faster
on raid 1+0 than
raid 5, but it shouldn't be HUGE.  When I contacted 3ware, they basically
said the same thing.
I do agree that writes are MUCH faster on 1+0 than raid 5.

Any thoughts?

L8r...

Matt


> On Tue, Nov 05, 2002 at 01:22:25PM -0700, Matt Heaton wrote:
> > each NFS server.  So even though our throughput of only 1.5 MB isn't
high.
> > The number of files per second is
> > actually quite high, and causes things to slow down because of seek time
> > issues.  PLEASE GIVE US CACHEFS SOMEONE??
>
> How is cachefs going to help?  The kernel is already trying to cache data
> as much as possible.  Once you're trying to serve more data than you have
> RAM, this are naturally going to degreate quite significantly as the
system
> becomes seek bound.
>
> > Does anyone have experience with IDE Raid arrays that get over 250 tps
in
> > iostat that work fine?  I would
> > be VERY VERY VERY interested to find out.
>
> Use raid1+0 and you'll be much happier, as read requests will be balanced
> over multiple drives (mirroring means the same data can be read from all
> of the mirrors).  Additionally, you'll have much lower CPU utilization
> and writes won't cause all disks in the array to seek for strip updates.
> Read the archives for the past couple of weeks for another example of the
> performance increase when going from raid5 to raid1+0.
>
> -ben
>
>




-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

RE: nfs performance problem

[Lever, Charles]

> Cachefs will help quite a lot in my opinion because it 
> doesn't just store
> the files in RAM,
> it uses the hard drive.  So if you have an NFS client with an 
> extra 5 gig
> that you can
> designate as cache then reads to the NFS server will go down 
> DRAMATICALLY as
> it will hit local cache on the NFS clients drive.

yes, that's true, until you consider that without something
like v4 delegation, the client still has to contact the server
to keep its cache up to date.

matt, you really do want a faster server in this case.


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Benjamin LaHaise]

You're wrong.  Reads (not sequential, but scattered) on a 1+0 setup will be 
faster as the raid1 driver optimizes requests somewhat, plus 2 (or more for 
larger raid1s of disks) drives will be able to service requests for the same 
stripe offset on different disks.  There is no way a raid5 can service two 
requests for the same stripe offset at different offsets in the array.  And 
even for a read-heavy workload, there are still writes to update metadata 
(and journal unless you've got a separate journal device), and the impact 
of those is that *all* reads to the array have to suffer from seeks when 
even the smallest write is active.

On a side note, make sure you have the filesystem mounted with the noatime 
flag if you can afford losing atimes.

		-ben

On Tue, Nov 05, 2002 at 01:46:04PM -0700, Matt Heaton wrote:
> Cachefs will help quite a lot in my opinion because it doesn't just store
> the files in RAM,
> it uses the hard drive.  So if you have an NFS client with an extra 5 gig
> that you can
> designate as cache then reads to the NFS server will go down DRAMATICALLY as
> it will hit local cache on the NFS clients drive.
> 
> I agree raid 1+0 should be much faster for writes and a little for read, but
> RAID 5 still
> reads from all drives simultaneously (Has to read parity in too I know), but
> can read
> all 7 drives at once instead of only 4 drives at once in a raid 1+0
> configuration with 8 drives
> in the array.  I have never used 1+0 so I am only talking about physical
> drive layout rather
> than any personal experience.  Are my assumptions correct that raid 5 does
> in fact read
> from all drives at the same time?  If so, reading might be a LITTLE faster
> on raid 1+0 than
> raid 5, but it shouldn't be HUGE.  When I contacted 3ware, they basically
> said the same thing.
> I do agree that writes are MUCH faster on 1+0 than raid 5.
> 
> Any thoughts?
> 
> L8r...
> 
> Matt
> 
> 
> > On Tue, Nov 05, 2002 at 01:22:25PM -0700, Matt Heaton wrote:
> > > each NFS server.  So even though our throughput of only 1.5 MB isn't
> high.
> > > The number of files per second is
> > > actually quite high, and causes things to slow down because of seek time
> > > issues.  PLEASE GIVE US CACHEFS SOMEONE??
> >
> > How is cachefs going to help?  The kernel is already trying to cache data
> > as much as possible.  Once you're trying to serve more data than you have
> > RAM, this are naturally going to degreate quite significantly as the
> system
> > becomes seek bound.
> >
> > > Does anyone have experience with IDE Raid arrays that get over 250 tps
> in
> > > iostat that work fine?  I would
> > > be VERY VERY VERY interested to find out.
> >
> > Use raid1+0 and you'll be much happier, as read requests will be balanced
> > over multiple drives (mirroring means the same data can be read from all
> > of the mirrors).  Additionally, you'll have much lower CPU utilization
> > and writes won't cause all disks in the array to seek for strip updates.
> > Read the archives for the past couple of weeks for another example of the
> > performance increase when going from raid5 to raid1+0.
> >
> > -ben
> >
> >
> 

-- 
"Do you seek knowledge in time travel?"


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

RE: nfs performance problem

[Lever, Charles]

> On a side note, make sure you have the filesystem mounted 
> with the noatime flag if you can afford losing atimes.

to be scrupulously clear, that's on the server side, not on
the client side.  the NFS client ignores the "noatime" mount
option, since file access timestamps are managed on the server.


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Ragnar Kjørstad]

On Tue, Nov 05, 2002 at 08:55:41PM +0100, poczta.dotcom.pl wrote:
> > How is local performance?
>=20
> local read/writes are ok
>=20
> > What filesystem do you use?
>=20
> reiserfs on top of  lvm (to be able to get snapshots)

What reiserfs-version? 3.5 or 3.6?=20
What on-disk format? 3.5 or 3.6?
There should be an entry in the kernel-log from when the filesystem is
mounted.=20


Reiserfs-3.5 has some known performance-problems related to NFS - that
may be your problem.


> > What's the output from "iostat -x -d -k 60 3"?
> >
> below is output from 'iostat -d 60 3'  (- my iostat does not take '-k'
> option,
> and '-d' gives me empty output) - mayby this is enough?

And not -x either?
iostat fr=E5 sysstat-4.0.3-2 (from RedHat) includes a "-x" option for
extended output. Most importantly, it will tell how much of the time
each device is busy.

It's unclear from your posts if this is a io-related problem or not.
Unnless you're running reiserfs-3.5 my guess is that it is the
IO-performance that is the problem. 3ware 7500 controllers have rather=20
poor performance on RAID5 - especially for writes.=20

If that's correct, then switching to a different RAID-level or replacing
the 3ware-controller should solve the problem. SCSI- or FC- RAIDS are
easily 10 or 20 times faster than the 3ware RAID for some types of
operations, and if you want something cheaper there is also different
types of IDE-RAIDs.



--=20
Ragnar Kj=F8rstad
Big Storage


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[myciel]

----- Original Message -----
From: "Ragnar Kj=F8rstad" <nfs@ragnark.vestdata.no>
To: "poczta.dotcom.pl" <myciel@dotcom.pl>
Cc: <nfs@lists.sourceforge.net>
Sent: Wednesday, November 06, 2002 12:32 AM
Subject: Re: [NFS] nfs performance problem


> On Tue, Nov 05, 2002 at 08:55:41PM +0100, poczta.dotcom.pl wrote:
> > > How is local performance?
> >
> > local read/writes are ok
> >
> > > What filesystem do you use?
> >
> > reiserfs on top of  lvm (to be able to get snapshots)
>
> What reiserfs-version? 3.5 or 3.6?
> What on-disk format? 3.5 or 3.6?

3.6.25

>
> > > What's the output from "iostat -x -d -k 60 3"?
> > >
> > below is output from 'iostat -d 60 3'  (- my iostat does not take '-k=
'
> > option,
> > and '-d' gives me empty output) - mayby this is enough?
>
> And not -x either?
now I have 4.0.6,
iostat -x -k 60 3  - outputs only column titles every 60 seconds  but lin=
e
where numbers should be is blank,
the same I have on other machine (fresh gentoo linux)
> iostat fr=E5 sysstat-4.0.3-2 (from RedHat) includes a "-x" option for
> extended output. Most importantly, it will tell how much of the time
> each device is busy.

anyway local writes are ok
>
> It's unclear from your posts if this is a io-related problem or not.
> Unnless you're running reiserfs-3.5 my guess is that it is the
> IO-performance that is the problem. 3ware 7500 controllers have rather
> poor performance on RAID5 - especially for writes.

reiserfs 3.6.25, raid 5,
ok, I can understand raid 5 is not fast but getting below 2Mbytes/s
is really poor :-(
>
> If that's correct, then switching to a different RAID-level or replacin=
g
> the 3ware-controller should solve the problem. SCSI- or FC- RAIDS are
> easily 10 or 20 times faster than the 3ware RAID for some types of
> operations, and if you want something cheaper there is also different
> types of IDE-RAIDs.
>

what kind of IDE-RAID would You suggest?


thanks

rafal mycielski



-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Ragnar Kjørstad]

On Wed, Nov 06, 2002 at 09:59:01AM +0100, myciel wrote:
> > > > What filesystem do you use?
> > >
> > > reiserfs on top of  lvm (to be able to get snapshots)
> >
> > What reiserfs-version? 3.5 or 3.6?
> > What on-disk format? 3.5 or 3.6?
>=20
> 3.6.25

That's the reiserfs-version.
The on-disk format must be either "3.5" or "3.6".
There should be a message at mount-time telling you wich one, or you can
use "debugreiserfs <device>" to check it.

> > Unnless you're running reiserfs-3.5 my guess is that it is the
> > IO-performance that is the problem. 3ware 7500 controllers have rathe=
r
> > poor performance on RAID5 - especially for writes.
>=20
> reiserfs 3.6.25, raid 5,
> ok, I can understand raid 5 is not fast but getting below 2Mbytes/s
> is really poor :-(

Well, updating a single byte is a very very expensive operation on
raid5. I agree that 2MB/s is worse than one should expect though, so
there may be something else going on.

The fact that you say local writes are faster could indicate that the
problem is not only io-related.=20

Maybe there is some packet-loss? That kills performance on nfs.
Is there anything in the kernel-log on the clients to indicate the
problem?

> > If that's correct, then switching to a different RAID-level or replac=
ing
> > the 3ware-controller should solve the problem. SCSI- or FC- RAIDS are
> > easily 10 or 20 times faster than the 3ware RAID for some types of
> > operations, and if you want something cheaper there is also different
> > types of IDE-RAIDs.
>=20
> what kind of IDE-RAID would You suggest?

A BigStorage IDE-RAID of course :)
I'll get back to you off-list about that.

But it's still not clear if your problems are raid-related or
network-related. Possible a combination.



--=20
Ragnar Kj=F8rstad
Big Storage


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[myciel]

> On Wed, Nov 06, 2002 at 09:59:01AM +0100, myciel wrote:
> > > > > What filesystem do you use?
> > > >
> > > > reiserfs on top of  lvm (to be able to get snapshots)
> > >
> > > What reiserfs-version? 3.5 or 3.6?
> > > What on-disk format? 3.5 or 3.6?
> >
> > 3.6.25
>
> That's the reiserfs-version.
> The on-disk format must be either "3.5" or "3.6".
> There should be a message at mount-time telling you wich one, or you can
> use "debugreiserfs <device>" to check it.
>

format 3.6 with standard journal

> > > Unnless you're running reiserfs-3.5 my guess is that it is the
> > > IO-performance that is the problem. 3ware 7500 controllers have rather
> > > poor performance on RAID5 - especially for writes.
> >
> > reiserfs 3.6.25, raid 5,
> > ok, I can understand raid 5 is not fast but getting below 2Mbytes/s
> > is really poor :-(
>
> Well, updating a single byte is a very very expensive operation on
> raid5. I agree that 2MB/s is worse than one should expect though, so
> there may be something else going on.
>
> The fact that you say local writes are faster could indicate that the
> problem is not only io-related.
>
> Maybe there is some packet-loss?
I don't see any packet losses - at least as I can see with ping big packets
like 16k or 32k

>That kills performance on nfs.
> Is there anything in the kernel-log on the clients to indicate the
> problem?
>

yesterday evening I switched from udp to tcp,
performace is much better but I get a lot of kernel messages in syslog:

Nov  6 12:40:45 intler kernel: RPC request reserved 272 but used 276
Nov  6 12:40:50 intler kernel: RPC request reserved 240 but used 244
Nov  6 12:40:54 intler kernel: RPC request reserved 244 but used 248

what does above mean?

> >
> > what kind of IDE-RAID would You suggest?
>
> A BigStorage IDE-RAID of course :)
> I'll get back to you off-list about that.

ok

rafal mycielski



-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[pwitting]

> From: =?iso-8859-1?Q?Ragnar_Kj=F8rstad?= <nfs@ragnark.vestdata.no>
> On Tue, Nov 05, 2002 at 08:55:41PM +0100, poczta.dotcom.pl wrote:
> > > How is local performance?

> > local read/writes are ok

> > > What filesystem do you use?

> > reiserfs on top of lvm (to be able to get snapshots)

I'm just going to throw my two cents in here. It sounds like the box is
getting a lot of random IO (~500k maildirs) which means he is most likely
seeing a LOT of seek issues, and RAID 5 is probably a really bad choice.
Cheap controllers and fast RAID 5 just don't go together, there's too much
math going on (the parity bit moves, remember) Matt, you really need to
trust the folks on the list on this and try it. You'll lose some capacity
(drop from 800GB to 480GB per array), but hopefully the performance boost
will be worth it.

Now, on the plus side, you're already using LVM, so I have a better idea
than RAID 1+0 for your situation. Build six RAID 1 arrays, then use LVM to
create a single logical volume out of the group. The randomness should
ensure that if your server receives 4 simultaneous read/write requests, they
will each have their own "spindle", an there will be no performance robbing
back and forth seeking, meaning the requests get handled faster, meaning
there's less chance for the system to start thrashing with seek requests.

I usually think of each RAID volume as a "virtual spindle", capable of
handling 1 request at a time. While each individual transaction might be
slower, (no striping to speed things up) there's six lanes to handle the
requests. Server thinking versus Workstation thinking.

Hopefully you can experiment with this, If this things production already it
may be too late.


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Baker, Byran]

We have a couple of NFS servers with 3ware cards.  One is a 3ware 7850 (8
Maxtor 160GB - 5400RPM) and one is a 3ware 7500-12, both configured with
RAID 5 + Hot Spare.  I have all the latest NFS patches applied.  If I access
an NFS mounted directory on one of these systems and copy a bunch of little
files (1-2KB), I get 150-250 tps.  If I access larger files (90% > 1MB each)
I will normally see 25-60MB/s transfer with 500-700 tps, with peaks of over
1000 tps - A good indication that disk seek is to blame for the problem with
the small files.

Thanks,
-Byran

From: "Matt Heaton" <admin@0catch.com>
To: "poczta.dotcom.pl" <myciel@dotcom.pl>,
	<nfs@lists.sourceforge.net>
Subject: Re: [NFS] nfs performance problem
Date: Tue, 5 Nov 2002 13:22:25 -0700

I have 3ware cards as well (2 7500 Series 8 drive), just like yours.  ALso
with 120 GIG drives (7200)
RPM.  Whever I get 150-200 tps in iostat then my NFS runs SO SLOW.  I also
only get 1 MB, to 1.5 MB
over NFS.  The local speed seems to be ok, not great though.   MY OPINION is
that this is because of seek time on the raid
array.   I am serving very small files, just like you are.  I am requesting
about 200-300 files per second from
each NFS server.  So even though our throughput of only 1.5 MB isn't high.
The number of files per second is
actually quite high, and causes things to slow down because of seek time
issues.  PLEASE GIVE US CACHEFS SOMEONE??

Does anyone have experience with IDE Raid arrays that get over 250 tps in
iostat that work fine?  I would
be VERY VERY VERY interested to find out.

L8r...

Matt


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Matt Heaton]

I didn't understand?  What is a good indication of poor seek time?  The high
number for TPS?  I thought a high number meant the drive was handling a huge
amount of transactions per second, and not necessarily slow.  I would LOVE
to
have my NFS servers push out 500 TPS on iostat.  I have 10 webservers that
each ask for between 10-25 files per second from our 3ware 7850 controllers.
We get about 200 TPS on the NFS server.  Much more and the clients become
VERY SLUGGISH, and I get blocked R processes in vmstat on the client.
Sometimes
20-40 blocked R processes in vmstat.  As soon as I offload some of files
from a busy
NFS server to another NFS server then the clients settle down and everything
works
great again, but I only get 1-1.5MB out of each NFS server before this
happens!
I can get really high TPS out of the server LOCALLY when I do copies etc,
its just
over NFS with a huge amount of really small files that everything stinks.

L8r...
Matt



> We have a couple of NFS servers with 3ware cards.  One is a 3ware 7850 (8
> Maxtor 160GB - 5400RPM) and one is a 3ware 7500-12, both configured with
> RAID 5 + Hot Spare.  I have all the latest NFS patches applied.  If I
access
> an NFS mounted directory on one of these systems and copy a bunch of
little
> files (1-2KB), I get 150-250 tps.  If I access larger files (90% > 1MB
each)
> I will normally see 25-60MB/s transfer with 500-700 tps, with peaks of
over
> 1000 tps - A good indication that disk seek is to blame for the problem
with
> the small files.
>
> Thanks,
> -Byran
>
> From: "Matt Heaton" <admin@0catch.com>
> To: "poczta.dotcom.pl" <myciel@dotcom.pl>,
> <nfs@lists.sourceforge.net>
> Subject: Re: [NFS] nfs performance problem
> Date: Tue, 5 Nov 2002 13:22:25 -0700
>
> I have 3ware cards as well (2 7500 Series 8 drive), just like yours.  ALso
> with 120 GIG drives (7200)
> RPM.  Whever I get 150-200 tps in iostat then my NFS runs SO SLOW.  I also
> only get 1 MB, to 1.5 MB
> over NFS.  The local speed seems to be ok, not great though.   MY OPINION
is
> that this is because of seek time on the raid
> array.   I am serving very small files, just like you are.  I am
requesting
> about 200-300 files per second from
> each NFS server.  So even though our throughput of only 1.5 MB isn't high.
> The number of files per second is
> actually quite high, and causes things to slow down because of seek time
> issues.  PLEASE GIVE US CACHEFS SOMEONE??
>
> Does anyone have experience with IDE Raid arrays that get over 250 tps in
> iostat that work fine?  I would
> be VERY VERY VERY interested to find out.
>
> L8r...
>
> Matt
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by: See the NEW Palm
> Tungsten T handheld. Power & Color in a compact size!
> http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
> _______________________________________________
> NFS maillist  -  NFS@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/nfs
>
>



-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Ragnar Kjørstad]

On Thu, Nov 07, 2002 at 09:19:44AM -0600, Baker, Byran wrote:
> We have a couple of NFS servers with 3ware cards.  One is a 3ware 7850 =
(8
> Maxtor 160GB - 5400RPM) and one is a 3ware 7500-12, both configured wit=
h
> RAID 5 + Hot Spare.  I have all the latest NFS patches applied.  If I a=
ccess
> an NFS mounted directory on one of these systems and copy a bunch of li=
ttle
> files (1-2KB), I get 150-250 tps.  If I access larger files (90% > 1MB =
each)
> I will normally see 25-60MB/s transfer with 500-700 tps, with peaks of =
over
> 1000 tps - A good indication that disk seek is to blame for the problem=
 with
> the small files.


Just to clarify here.

I bet when you transfer 1-2KB files you are able to transfer 150-250
files pr second, right?=20

When you transfer larger files the files pr second rate goes down, but
the tps-rate increases because each file is split into multiple scsi
requests.=20


And yes, I would agree with your conclution that disk seek is to blame.

Are those numbers for reads or writes? I would expect to see the same
tendency, but with stronger effect, for writes than reads.



--=20
Ragnar Kj=F8rstad
Big Storage


-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm 
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

nfs performance problem

[Andreas Schuldei]

Hi!

I need to tune a nfs server and client. on the server we have
several Tbyte of ~2Mbyte files and we need to transfer them read
only to the client. latency and throughput are crucial.

What nfs server should i use? i started with the
nfs-kernel-server on top of a kernel 2.6.22 on debian on the
server side. the client is a debian etch server (2.6.18 kernel)
with 1Gbyte e1000 intel network driver. later on we consider two
network cards on both machines to transfer 2Gbit/s. Jumboframes
are an option (how much will they help?)

Right now i have only four disks in the server and i get 50Mbyte
out of each of them, simultaniously, for real world loads (random
reads across the disk, trying to minimizing the seeks by reading
the files in one go with

for i in a b h i ; do ( find /var/disks/sd$i -type f | xargs -I=B0 dd if=3D=
=B0 bs=3D2M of=3D/dev/null status=3Dnoxfer 2>/dev/null & ) ; done

so with this (4*50 Mbyte/s) i should be able to saturate both
network cards.

accessing the disks with apache2-mpm-worker we get ~90Mbyte/s out
of the server, partly with considerable latency in the order of
magnitude of 10s.

I was hoping to get at least the same performance with much
better latency with nfs.

on the server i start 128 nfs servers (RPCNFSDCOUNT=3D128) and export
the disks like this:

/usr/sbin/exportfs -v
/var/disks/sda <world>(ro,async,wdelay,root_squash,no_subtree_check,anonuid=
=3D65534,anongid=3D65534)
/var/disks/sdb <world>(ro,async,wdelay,root_squash,no_subtree_check,anonuid=
=3D65534,anongid=3D65534)
/var/disks/sdh <world>(ro,async,wdelay,root_squash,no_subtree_check,anonuid=
=3D65534,anongid=3D65534)
/var/disks/sdi <world>(ro,async,wdelay,root_squash,no_subtree_check,anonuid=
=3D65534,anongid=3D65534)


on the client i mount them like this:

lotta:/var/disks/sda on /var/disks/sda type nfs (ro,hard,intr,proto=3Dtcp,r=
size=3D32k,addr=3D217.213.5.44)
lotta:/var/disks/sdb on /var/disks/sdb type nfs (ro,hard,intr,proto=3Dtcp,r=
size=3D32k,addr=3D217.213.5.44)
lotta:/var/disks/sdh on /var/disks/sdh type nfs (ro,hard,intr,proto=3Dtcp,r=
size=3D32k,addr=3D217.213.5.44)
lotta:/var/disks/sdi on /var/disks/sdi type nfs (ro,hard,intr,proto=3Dtcp,r=
size=3D32k,addr=3D217.213.5.44)

but when i then do the same dd again on the client i get
disappointing 60-70Mbyte/s altogether. from a single disk i get
~25Mbytes/s on the client side.

i played with some buffers /proc/sys/net/core/rmem_max and
/proc/sys/net/core/rmem_default and increased them to 256M on the
client.

i was suspecting that the nfs server reads the files in too small
chunks and tried to help it with =


for i in a  h i  ; do ( echo $((1024*6)) > /sys/block/sd$i/queue/read_ahead=
_kb ) ; done

to get it to read in the files in one go.

I would hope to at least double the speed. do
you have a benchmark tool that can tell me the latency? i tried
iozone and tried forcing it to only do read tests and did not get
any helpfull error or output at all. =



on the server:

nfsstat
Server rpc stats:
calls      badcalls   badauth    badclnt    xdrcall
98188885   0          0          0          0

Server nfs v3:
null         getattr      setattr      lookup       access       readlink
5599      0% 318417    0% 160       0% 132643    0% 227130    0% 0         =
0%
read         write        create       mkdir        symlink      mknod
97256921 99% 118313    0% 168       0% 0         0% 0         0% 0         =
0%
remove       rmdir        rename       link         readdir      readdirplus
162       0% 0         0% 0         0% 0         0% 0         0% 105556    =
0%
fsstat       fsinfo       pathconf     commit
0         0% 1270      0% 0         0% 7153      0%



cat /proc/net/rpc/nfsd
rc 0 118803 98069945
fh 0 0 0 0 0
io 3253902194 38428672
th 128 10156908 1462.848 365.212 302.100 252.204 311.632 187.508 142.708 14=
2.132 198.168 648.640
ra 256 97097262 0 0 0 0 0 0 0 0 0 64684
net 98188985 16 98188854 5619
rpc 98188885 0 0 0 0
proc2 18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
proc3 22 5599 318417 160 132643 227130 0 97256921 118313 168 0 0 0 162 0 0 =
0 0 105556 0 1270 0 7153
proc4 2 0 0
proc4ops 40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0=
 0 0 0 0 0 0 0 0






-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Bernd Schubert]

Hello Andreas,

On Thursday 25 October 2007 15:10:29 Andreas Schuldei wrote:
> lotta:/var/disks/sda on /var/disks/sda type nfs
> (ro,hard,intr,proto=tcp,rsize=32k,addr=217.213.5.44) lotta:/var/disks/sdb

try to increase rsize and wsize as much as possible, the maximum can be 
adjusted in /proc/fs/nfsd/max_block_size on the nfs server.

>
> but when i then do the same dd again on the client i get
> disappointing 60-70Mbyte/s altogether. from a single disk i get
> ~25Mbytes/s on the client side.
>
> i played with some buffers /proc/sys/net/core/rmem_max and
> /proc/sys/net/core/rmem_default and increased them to 256M on the
> client.
>
> i was suspecting that the nfs server reads the files in too small
> chunks and tried to help it with
>
> for i in a  h i  ; do ( echo $((1024*6)) >
> /sys/block/sd$i/queue/read_ahead_kb ) ; done

Are your partitions on lvm or md? This wouldn't help then, AFAIK for lvm you 
can't do it via /sys, but must always use blockdev

>
> to get it to read in the files in one go.
>
> I would hope to at least double the speed. do
> you have a benchmark tool that can tell me the latency? i tried
> iozone and tried forcing it to only do read tests and did not get
> any helpfull error or output at all.

First of all, what is you *local* performance?


Cheers,
Bernd

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Andreas Schuldei]

* Bernd Schubert (bernd-schubert@gmx.de) [071025 15:54]:
> Hello Andreas,
> 
> On Thursday 25 October 2007 15:10:29 Andreas Schuldei wrote:
> > lotta:/var/disks/sda on /var/disks/sda type nfs
> > (ro,hard,intr,proto=tcp,rsize=32k,addr=217.213.5.44) lotta:/var/disks/sdb
> 
> try to increase rsize and wsize as much as possible, the maximum can be 
> adjusted in /proc/fs/nfsd/max_block_size on the nfs server.

the with a max_block_size and rsize=512k i did not achive a noticable performance impact.
increasing it to 2M decreased the performance to 12-12Mbytes/s per disk. do i need to adjust more 
tuneables to benefit from the increased size?


> > i was suspecting that the nfs server reads the files in too small
> > chunks and tried to help it with
> >
> > for i in a  h i  ; do ( echo $((1024*6)) >
> > /sys/block/sd$i/queue/read_ahead_kb ) ; done
> 
> Are your partitions on lvm or md? This wouldn't help then, AFAIK for lvm you 
> can't do it via /sys, but must always use blockdev

they are on xfs directly on disk, no lvm or md involved.

> > to get it to read in the files in one go.
> >
> > I would hope to at least double the speed. do
> > you have a benchmark tool that can tell me the latency? i tried
> > iozone and tried forcing it to only do read tests and did not get
> > any helpfull error or output at all.
> 
> First of all, what is you *local* performance?

i mentioned that i get 50Mbyte/s out of each disk, even
when doing the dd on all four disks at the same time.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Chuck Lever]

On Oct 25, 2007, at 9:10 AM, Andreas Schuldei wrote:
> Hi!
>
> I need to tune a nfs server and client. on the server we have
> several Tbyte of ~2Mbyte files and we need to transfer them read
> only to the client. latency and throughput are crucial.
>
> What nfs server should i use? i started with the
> nfs-kernel-server on top of a kernel 2.6.22 on debian on the
> server side. the client is a debian etch server (2.6.18 kernel)
> with 1Gbyte e1000 intel network driver. later on we consider two
> network cards on both machines to transfer 2Gbit/s. Jumboframes
> are an option (how much will they help?)
>
> Right now i have only four disks in the server and i get 50Mbyte
> out of each of them, simultaniously, for real world loads (random
> reads across the disk, trying to minimizing the seeks by reading
> the files in one go with
>
> for i in a b h i ; do ( find /var/disks/sd$i -type f | xargs -I=B0 dd  =

> if=3D=B0 bs=3D2M of=3D/dev/null status=3Dnoxfer 2>/dev/null & ) ; done
>
> so with this (4*50 Mbyte/s) i should be able to saturate both
> network cards.
>
> accessing the disks with apache2-mpm-worker we get ~90Mbyte/s out
> of the server, partly with considerable latency in the order of
> magnitude of 10s.
>
> I was hoping to get at least the same performance with much
> better latency with nfs.

With a single client, you should not expect to get any better  =

performance than by running the web service on the NFS server.  The  =

advantage of using NFS under a web service is that you can  =

transparently scale horizontally.  When you add a second or third web  =

server that serves the same file set, you will see an effective  =

increase in the size of the data cache between your NFS server's  =

disks and the web servers.

But don't expect to get better data throughput over NFS than you see  =

on your local NFS server.  If anything, the 10s latency you see when  =

the web server is on the same system with the disks is indicative of  =

local file system configuration issues.

> on the server i start 128 nfs servers (RPCNFSDCOUNT=3D128) and export
> the disks like this:
>
> /usr/sbin/exportfs -v
> /var/disks/sda <world> =

> (ro,async,wdelay,root_squash,no_subtree_check,anonuid=3D65534,anongid=3D6=
5 =

> 534)
> /var/disks/sdb <world> =

> (ro,async,wdelay,root_squash,no_subtree_check,anonuid=3D65534,anongid=3D6=
5 =

> 534)
> /var/disks/sdh <world> =

> (ro,async,wdelay,root_squash,no_subtree_check,anonuid=3D65534,anongid=3D6=
5 =

> 534)
> /var/disks/sdi <world> =

> (ro,async,wdelay,root_squash,no_subtree_check,anonuid=3D65534,anongid=3D6=
5 =

> 534)

On the server, mounting the web data file systems with "noatime" may  =

help reduce the number of seeks on the disks.

Also, the "async" export option won't have any effect on reads.

Check your block device configuration as well.  You may find that  =

varying the RAID configuration, file system type (ext3 v. xfs) and  =

stripe/chunk size could impact your server's performance.  You might  =

find that the deadline disk scheduler performs a little better than  =

the default cfq scheduler.

It goes without saying that you should make sure your disk subsystem  =

is healthy.  I've found that, for example, SATA drives in hot-swap  =

enclosures are sometimes affected by silent SATA transport errors  =

that result in slow performance.  Check dmesg carefully to ensure you  =

are getting the highest possible speed settings.  If only one of your  =

drives is running significantly slower than the others, it will have  =

a significant impact on the performance of a RAID group.

> on the client i mount them like this:
>
> lotta:/var/disks/sda on /var/disks/sda type nfs  =

> (ro,hard,intr,proto=3Dtcp,rsize=3D32k,addr=3D217.213.5.44)
> lotta:/var/disks/sdb on /var/disks/sdb type nfs  =

> (ro,hard,intr,proto=3Dtcp,rsize=3D32k,addr=3D217.213.5.44)
> lotta:/var/disks/sdh on /var/disks/sdh type nfs  =

> (ro,hard,intr,proto=3Dtcp,rsize=3D32k,addr=3D217.213.5.44)
> lotta:/var/disks/sdi on /var/disks/sdi type nfs  =

> (ro,hard,intr,proto=3Dtcp,rsize=3D32k,addr=3D217.213.5.44)

There are some client-side mount options that might also help.  Using  =

"nocto" and "actimeo=3D7200" could reduce synchonous NFS protocol  =

overhead.  I also notice a significant amount of readdirplus  =

traffic.  Readdirplus requests are fairly heavyweight, and in this  =

scenario may be unneeded overhead.  Your client might support the  =

recently added "nordirplus" mount option, which could be helpful.

I wonder if "rsize=3D32k" is supported - you might want "rsize=3D32768"  =

instead.  Or better, let the client and server negotiate the maximum  =

that each supports automatically by leaving this option off.  You can  =

check what options are in effect on each NFS mount point by looking  =

in /proc/self/mountstats on the client.

Enabling jumbo frames between your NFS server and client will help.   =

Depending on your NIC, though, it may introduce some instability  =

(driver and hardware mileage may vary).

Since you currently have only one client, you might consider running  =

the client and server back-to-back (ie replace any hub or switch with  =

a simple cross-over link) to eliminate extra network overhead.   =

Getting a high performance switch when you add more clients is key to  =

making this configuration scale well -- a $99 special won't cut it.

> but when i then do the same dd again on the client i get
> disappointing 60-70Mbyte/s altogether. from a single disk i get
> ~25Mbytes/s on the client side.

25MB/s is fairly typical for Linux NFS servers.

> i played with some buffers /proc/sys/net/core/rmem_max and
> /proc/sys/net/core/rmem_default and increased them to 256M on the
> client.

You should consider similar network tuning on the server.  Use a  =

network benchmarking tool like iperf to assist.

> i was suspecting that the nfs server reads the files in too small
> chunks and tried to help it with
>
> for i in a  h i  ; do ( echo $((1024*6)) > /sys/block/sd$i/queue/ =

> read_ahead_kb ) ; done
>
> to get it to read in the files in one go.

Insufficient read-ahead on your server may be an issue here.  Read  =

traffic from the client often arrives at the server out of order,  =

preventing the server from cleanly detecting sequential reads.  I  =

believe there was a recent change to the NFS server that addresses  =

this issue.

> I would hope to at least double the speed.

IMO you can do that only by adding more clients.

> do
> you have a benchmark tool that can tell me the latency? i tried
> iozone and tried forcing it to only do read tests and did not get
> any helpfull error or output at all.

Use "iozone -a -i 1" to run read tests.  You can narrow the test down  =

to 2MB sequential reads if you want.  Take a look at "iozone -h"  =

output for more details.

> on the server:
>
> nfsstat
> Server rpc stats:
> calls      badcalls   badauth    badclnt    xdrcall
> 98188885   0          0          0          0
>
> Server nfs v3:
> null         getattr      setattr      lookup       access        =

> readlink
> 5599      0% 318417    0% 160       0% 132643    0% 227130    0%  =

> 0         0%
> read         write        create       mkdir        symlink      mknod
> 97256921 99% 118313    0% 168       0% 0         0% 0         0%  =

> 0         0%
> remove       rmdir        rename       link         readdir       =

> readdirplus
> 162       0% 0         0% 0         0% 0         0% 0         0%  =

> 105556    0%
> fsstat       fsinfo       pathconf     commit
> 0         0% 1270      0% 0         0% 7153      0%
>
>
>
> cat /proc/net/rpc/nfsd
> rc 0 118803 98069945
> fh 0 0 0 0 0
> io 3253902194 38428672
> th 128 10156908 1462.848 365.212 302.100 252.204 311.632 187.508  =

> 142.708 142.132 198.168 648.640
> ra 256 97097262 0 0 0 0 0 0 0 0 0 64684
> net 98188985 16 98188854 5619
> rpc 98188885 0 0 0 0
> proc2 18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> proc3 22 5599 318417 160 132643 227130 0 97256921 118313 168 0 0 0  =

> 162 0 0 0 0 105556 0 1270 0 7153
> proc4 2 0 0
> proc4ops 40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0  =

> 0 0 0 0 0 0 0 0 0 0 0 0

--
Chuck Lever
chuck[dot]lever[at]oracle[dot]com




-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Andreas Schuldei]

* Chuck Lever (chuck.lever@oracle.com) [071025 20:25]:
> On Oct 25, 2007, at 9:10 AM, Andreas Schuldei wrote:
> >Hi!
> >
> >I need to tune a nfs server and client. on the server we have
> >several Tbyte of ~2Mbyte files and we need to transfer them read
> >only to the client. latency and throughput are crucial.

Because i have Tbytes of data but only a few Gbytes or RAM my
cache hits are rather unlikely. =


> >Right now i have only four disks in the server and i get 50Mbyte
> >out of each of them, simultaniously, for real world loads (random
> >reads across the disk, trying to minimizing the seeks by reading
> >the files in one go with
> >
> >for i in a b h i ; do ( find /var/disks/sd$i -type f | xargs -I=B0 dd if=
=3D=B0 bs=3D2M of=3D/dev/null status=3Dnoxfer =

> >2>/dev/null & ) ; done
> >
> >so with this (4*50 Mbyte/s) i should be able to saturate both
> >network cards.

note that this is my server's disk io performance.

> With a single client, you should not expect to get any better performance=
 than by running the web service on the NFS =

> server.  The advantage of using NFS under a web service is that you can t=
ransparently scale horizontally.  When you add =

> a second or third web server that serves the same file set, you will see =
an effective increase in the size of the data =

> cache between your NFS server's disks and the web servers.

Not with terabyte of data and a distributed access pattern.
Certainly i will have some cache hits but not enough to be able
to serv considerable amounts out of RAM.


> But don't expect to get better data throughput over NFS than you see on y=
our local NFS server. =


That is exactly the point. on my server i get 4*50Mbytes =3D
200Mbyte/s out of the disks (with the above FOR loop around the
find and dd) and when i export on the same server the disks to an
nfs client i all of a sudden loose ~75% of the performance.

> If anything, the 10s =

> latency you see when the web server is on the same system with the disks =
is indicative of local file system =

> configuration issues.

how can i measure the latency on the local machine? i would be
very interested in seeing how it behaves latency wise.

> >on the server i start 128 nfs servers (RPCNFSDCOUNT=3D128) and export
> >the disks like this:
> >
> >/usr/sbin/exportfs -v
> >/var/disks/sda <world>(ro,async,wdelay,root_squash,no_subtree_check,anon=
uid=3D65534,anongid=3D65534)
> >/var/disks/sdb <world>(ro,async,wdelay,root_squash,no_subtree_check,anon=
uid=3D65534,anongid=3D65534)
> >/var/disks/sdh <world>(ro,async,wdelay,root_squash,no_subtree_check,anon=
uid=3D65534,anongid=3D65534)
> >/var/disks/sdi <world>(ro,async,wdelay,root_squash,no_subtree_check,anon=
uid=3D65534,anongid=3D65534)
> =

> On the server, mounting the web data file systems with "noatime" may help=
 reduce the number of seeks on the disks.

yes, we do that already.

> >on the client i mount them like this:
> >
> >lotta:/var/disks/sda on /var/disks/sda type nfs (ro,hard,intr,proto=3Dtc=
p,rsize=3D32k,addr=3D217.213.5.44)
> >lotta:/var/disks/sdb on /var/disks/sdb type nfs (ro,hard,intr,proto=3Dtc=
p,rsize=3D32k,addr=3D217.213.5.44)
> >lotta:/var/disks/sdh on /var/disks/sdh type nfs (ro,hard,intr,proto=3Dtc=
p,rsize=3D32k,addr=3D217.213.5.44)
> >lotta:/var/disks/sdi on /var/disks/sdi type nfs (ro,hard,intr,proto=3Dtc=
p,rsize=3D32k,addr=3D217.213.5.44)
> =

> There are some client-side mount options that might also help.  Using "no=
cto" and "actimeo=3D7200" could reduce =

> synchonous NFS protocol overhead.  I also notice a significant amount of =
readdirplus traffic.  Readdirplus requests are =

> fairly heavyweight, and in this scenario may be unneeded overhead.  Your =
client might support the recently added =

> "nordirplus" mount option, which could be helpful.
> =

> I wonder if "rsize=3D32k" is supported - you might want "rsize=3D32768" i=
nstead.

i think that gave an effect. now i am in the 90-100Mbyte/s
ballpark and might hit the one-nic (1gbit) bottleneck.

> Or better, let the client and server =

> negotiate the maximum that each supports automatically by leaving this op=
tion off.  You can check what options are in =

> effect on each NFS mount point by looking in /proc/self/mountstats on the=
 client.

there it says now, after i specified rsize=3D2097152:
        opts:   rw,vers=3D3,rsize=3D1048576,wsize=3D1048576,acregmin=3D3,ac=
regmax=3D60,acdirmin=3D30,acdirmax=3D60,hard,intr,nolock,proto=3Dtcp,timeo=
=3D600,retrans=3D2,sec=3Dsys

i am surprised that it did not protest when it could not parse
the "k". note that it it only took 1M chunks. how come?

> Enabling jumbo frames between your NFS server and client will help.  Depe=
nding on your NIC, though, it may introduce =

> some instability (driver and hardware mileage may vary).

i will test that and bonding two nicks.

> Insufficient read-ahead on your server may be an issue here.  Read traffi=
c from the client often arrives at the server =

> out of order, preventing the server from cleanly detecting sequential rea=
ds.  I believe there was a recent change to =

> the NFS server that addresses this issue.

when did that go in? do i need to activate that somehow?

how can i measure the latency on a loaded server? both locally
and over nfs?

/andreas

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Chuck Lever]

[-- Attachment #1: Type: text/plain, Size: 4328 bytes --]

Andreas Schuldei wrote:
> * Chuck Lever (chuck.lever@oracle.com) [071025 20:25]:
>> On Oct 25, 2007, at 9:10 AM, Andreas Schuldei wrote:
>> With a single client, you should not expect to get any better performance than by running the web service on the NFS 
>> > server.  The advantage of using NFS under a web service is that you can transparently scale horizontally.  When you add 
>> > a second or third web server that serves the same file set, you will see an effective increase in the size of the data 
>> > cache between your NFS server's disks and the web servers.
> 
> Not with terabyte of data and a distributed access pattern.
> Certainly i will have some cache hits but not enough to be able
> to serv considerable amounts out of RAM.

If you don't intend to scale horizontally by adding more clients, then 
inserting NFS between your disks and your web service is not recommended 
unless you have security requirement that needs to be solved by 
administering your data separate from running the web service.

>> If anything, the 10s 
>> latency you see when the web server is on the same system with the disks is indicative of local file system 
>> configuration issues.
> 
> how can i measure the latency on the local machine? i would be
> very interested in seeing how it behaves latency wise.

The iostat command has options for displaying disk service time.

>>> on the client i mount them like this:
>>>
>>> lotta:/var/disks/sda on /var/disks/sda type nfs (ro,hard,intr,proto=tcp,rsize=32k,addr=217.213.5.44)
>>> lotta:/var/disks/sdb on /var/disks/sdb type nfs (ro,hard,intr,proto=tcp,rsize=32k,addr=217.213.5.44)
>>> lotta:/var/disks/sdh on /var/disks/sdh type nfs (ro,hard,intr,proto=tcp,rsize=32k,addr=217.213.5.44)
>>> lotta:/var/disks/sdi on /var/disks/sdi type nfs (ro,hard,intr,proto=tcp,rsize=32k,addr=217.213.5.44)
>> There are some client-side mount options that might also help.  Using "nocto" and "actimeo=7200" could reduce 
>> synchonous NFS protocol overhead.  I also notice a significant amount of readdirplus traffic.  Readdirplus requests are 
>> fairly heavyweight, and in this scenario may be unneeded overhead.  Your client might support the recently added 
>> "nordirplus" mount option, which could be helpful.
>>
>> I wonder if "rsize=32k" is supported - you might want "rsize=32768" instead.
> 
> i think that gave an effect. now i am in the 90-100Mbyte/s
> ballpark and might hit the one-nic (1gbit) bottleneck.
> 
>> Or better, let the client and server 
>> negotiate the maximum that each supports automatically by leaving this option off.  You can check what options are in 
>> effect on each NFS mount point by looking in /proc/self/mountstats on the client.
> 
> there it says now, after i specified rsize=2097152:
>         opts:   rw,vers=3,rsize=1048576,wsize=1048576,acregmin=3,acregmax=60,acdirmin=30,acdirmax=60,hard,intr,nolock,proto=tcp,timeo=600,retrans=2,sec=sys
> 
> i am surprised that it did not protest when it could not parse
> the "k". note that it it only took 1M chunks. how come?

1MB is the maximum that both your server and client support.  Again, if 
you don't specify rsize at all, the maximum both support will be 
negotiated automatically.

>> Insufficient read-ahead on your server may be an issue here.  Read traffic from the client often arrives at the server 
>> out of order, preventing the server from cleanly detecting sequential reads.  I believe there was a recent change to 
>> the NFS server that addresses this issue.
> 
> when did that go in? do i need to activate that somehow?

I don't think activation is needed.  Someone else on the list can speak 
to when it was included in recent kernels or in distributions.  However, 
since you are already close to your network limits, read-ahead is 
probably not an issue for you.

> how can i measure the latency on a loaded server? both locally
> and over nfs?

The iostat command, as mentioned above, will report on local disk 
activity.  NFS client activity has not been integrated into iostat, but 
there are some NFS client metrics available in /proc/self/mountstats.  I 
have a pair of Python tools that can extract and display RPC request 
latency information.  See:

   http://oss.oracle.com/~cel/linux-2.6/2.6.21/iostat-ms

and

   http://oss.oracle.com/~cel/linux-2.6/2.6.21/mountstats

[-- Attachment #2: chuck.lever.vcf --]
[-- Type: text/x-vcard, Size: 248 bytes --]

begin:vcard
fn:Chuck Lever
n:Lever;Chuck
org:Oracle Corporation;Corporate Architecture, Linux Projects Group
email;internet:chuck.lever@oracle.com
title:Principal Member of Staff
tel;work:+1 248 614 5091
x-mozilla-html:FALSE
version:2.1
end:vcard


[-- Attachment #3: Type: text/plain, Size: 314 bytes --]

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/

[-- Attachment #4: Type: text/plain, Size: 140 bytes --]

_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Talpey, Thomas]

At 03:34 PM 10/25/2007, Andreas Schuldei wrote:
>> But don't expect to get better data throughput over NFS than you see 
>on your local NFS server. 
>
>That is exactly the point. on my server i get 4*50Mbytes =
>200Mbyte/s out of the disks (with the above FOR loop around the
>find and dd) and when i export on the same server the disks to an
>nfs client i all of a sudden loose ~75% of the performance.

Andreas - you mentioned earlier in the thread that you're using 2.6.22
as the server kernel. You might consider moving it forward to 2.6.23,
as there is a significant readahead improvement/fix there. When you
run a local read, all readahead is perfect. However on older kernels
serving NFS, the reordering of reads due to Linux's multithreaded nfsd's
triggers several issues which spook the readahead into truly awful
behavior.

Fengguang Wu's changes are merged in 2.6.23, in testing with NFS/RDMA
we now see the same performance over the wire as locally, for any
number of readers.

My one caveat to this is that your files are relatively small, so I would
not expect a truly dramatic speedup from it. It's certainly worth trying,
there are other improvements in 2.6.23.

Do you have Infiniband or iWARP adapters in your environment? If you're
worried about bandwidth and application startup latency, using it could
certainly help. You'd need to run a 2.6.24-rc1 client though, and some
assembly is still required to use the server (it's not yet upstream).

Tom.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[dean hildebrand]

>... You might consider moving it forward to 2.6.23,
> as there is a significant readahead improvement/fix there.
>...
> Fengguang Wu's changes are merged in 2.6.23,
>...Tom,

Improving out-of-order nfsd requests is a great idea.  Is there a
description of Fengguang Wu's improvements or maybe a link to the
specific patches. I tried searching through git and such and not sure
if I found the right set of patches.

Dean


>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Splunk Inc.
> Still grepping through log files to find problems?  Stop.
> Now Search log events and configuration files using AJAX and a browser.
> Download your FREE copy of Splunk now >> http://get.splunk.com/
> _______________________________________________
> NFS maillist  -  NFS@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/nfs
>

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Andreas Schuldei]

* Bernd Schubert (bernd-schubert@gmx.de) [071027 01:39]:
> Hello Andreas,
> 
> On Thursday 25 October 2007 15:10:29 Andreas Schuldei wrote:
> > lotta:/var/disks/sda on /var/disks/sda type nfs
> > (ro,hard,intr,proto=tcp,rsize=32k,addr=217.213.5.44) lotta:/var/disks/sdb
> 
> try to increase rsize and wsize as much as possible, the maximum can be 
> adjusted in /proc/fs/nfsd/max_block_size on the nfs server.

i cant increase this to more then 1M. is that the hard limit?



-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: nfs performance problem

[Talpey, Thomas]

At 09:35 PM 10/26/2007, dean hildebrand wrote:
>>... You might consider moving it forward to 2.6.23,
>> as there is a significant readahead improvement/fix there.
>>...
>> Fengguang Wu's changes are merged in 2.6.23,
>>...Tom,
>
>Improving out-of-order nfsd requests is a great idea.  Is there a
>description of Fengguang Wu's improvements or maybe a link to the
>specific patches. I tried searching through git and such and not sure
>if I found the right set of patches.

There are quite a few of them. A big one with lots of background info
is in mm/readahead.c, commit 122a21d11cbfda6d1e33cbc8ae9e4c4ee2f1886e

Check out the "readahead thrashing" results at the end of the log, especially
the comment "the more overall read density, the more possible gain", i.e. it's
the density and not the actual order which now drives the readahead.
Good stuff.

Tom.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs