zfs discuss - Mar 2010 - pool use from network poor performance

Hi

i have now two pools

rpool 2-way mirror  ( pata )
data 4-way raidz2   ( sata )

if i access to datapool from network , smb , nfs , ftp , sftp , jne...
i get only max 200 KB/s speeds
compared to rpool that give XX MB/S speeds to and from network it is slow.

Any ideas what reasons might be and how try to find reason.

Locally datapool works reasonable fast for me. 
# date && mkfile 1G testfile && date
Tuesday, March 23, 2010 07:52:19 AM EET
Tuesday, March 23, 2010 07:52:36 AM EET


Some information about system.
# cat /etc/release
                       OpenSolaris Development snv_134 X86
           Copyright 2010 Sun Microsystems, Inc.  All Rights Reserved.
                        Use is subject to license terms.
                             Assembled 01 March 2010

# isainfo -v
64-bit amd64 applications
        ahf sse3 sse2 sse fxsr amd_3dnowx amd_3dnow amd_mmx mmx cmov amd_sysc
        cx8 tsc fpu
32-bit i386 applications
        ahf sse3 sse2 sse fxsr amd_3dnowx amd_3dnow amd_mmx mmx cmov amd_sysc
        cx8 tsc fpu

thanks
-- 
This message posted from opensolaris.org

On Mon, Mar 22, 2010 at 10:58:05PM -0700, homerun wrote:
> if i access to datapool from network , smb , nfs , ftp , sftp , jne...
> i get only max 200 KB/s speeds
> compared to rpool that give XX MB/S speeds to and from network it is slow.
> 
> Any ideas what reasons might be and how try to find reason.
Maybe a shared interrupt between the sata controller and the network card, 
with devices or drivers that don''t play well with others.

--
Dan.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100323/dce3bd0a/attachment.bin>

Hi

Well what is changed in system.
replaced 4 sata disks with new and bigger disks.
same time recreated raidz to raidz2
updated OS from b132 to 134

It used to work with old setup.
Has there been some driver changes.
-- 
This message posted from opensolaris.org

what does prstat show?

We had a lot of trouble here using iscsi and zvols due to the cpu capping out
with speeds less than 20mb/sec.

After simply switching to Qlogic fibre HBAs and a file backed lu we went to
160mb/sec on that same test platform.
-- 
This message posted from opensolaris.org

Hi

Here is more specs.

MB:
K8N4-E SE 
- AMD Socket 754 CPU
- NVIDIA? nForce? 4 4X 
- PCI Express Architecture
- Gigabit LAN
- 4 SATA RAID Ports
- 10 USB2.0 Ports
http://www.asus.com/product.aspx?P_ID=TBx7PakpparxrK89&templete=2

Now situation is this :
with ftp : 
i can upload to datapool with speed ~45MB/s
download from datapool only with speed ~ 750 KB/s

So it is now read performance that is a problem.
Could it really be that nvidia network and sata drivers now share same IRQ and
that''s why performance is slow.
Mar 23 19:35:01 <hostname> unix: [ID 954099 kern.info] NOTICE: IRQ20 is
being shared by drivers with different interrupt levels.

This is just odd as this issue come when only changed pool physical disks and
also changed raidz to raidz2 and also update to build 134
-- 
This message posted from opensolaris.org

Hi

Instaled pci addon network card and disabled nvidia onboarad network.
All started to work as should.
So nge and nv_sata drivers does not work in b134 when shared IRQ is used.
-- 
This message posted from opensolaris.org

Awsome, glad to hear that you got it figured out. :)
-- 
This message posted from opensolaris.org