> From: zfs-discuss-boun...@opensolaris.org [mailto:zfs-discuss-
> boun...@opensolaris.org] On Behalf Of Eugen Leitl
> I have a pool of 8x ST31000340AS on an LSI 8-port adapter as
> a raidz3 (no compression nor dedup) with reasonable bonnie++
> 1.03 values, e.g.  145 MByte/s Seq-Write @ 48% CPU and 291 MByte/s
> Seq-Read @ 53% CPU. 

For 8-disk raidz3 (effectively 5 disks) I would expect approx 640MB/s for both 
seq read and seq write.  The first halving (from 640 down to 291) could maybe 
be explained by bottlenecking through a single HBA or something like that, so I 
wouldn't be too concerned about that.  But the second halving, from 291 down to 
145 ... A single disk should do 128MB/sec no problem, so the whole pool writing 
at only 145MB/sec sounds wrong to me.

But as you said ... This isn't the area of complaint...  Moving on, you can 
start a new discussion about this if you want to later...

> My problem is pretty poor network throughput. An NFS
> mount on 12.04 64 bit Ubuntu (mtu 9000) or CIFS are
> read at about 23 MBytes/s. Windows 7 64 bit (also jumbo
> frames) reads at about 65 MBytes/s. The highest transfer
> speed on Windows just touches 90 MByte/s, before falling
> back to the usual 60-70 MBytes/s.
> Does anyone have any suggestions on how to debug/optimize
> throughput?

The first thing I would do is build another openindiana box and try NFS / CIFS 
to/from it.  See how it behaves.  Whenever I've seen this sort of problem 
before, it was version incompatibility requiring tweaks between the client and 
server.  I don't know which version of samba / solaris cifs is being used ... 
But at some point in history (win7), windows transitioned from NTLM v1 to v2, 
and at that point, all the older servers became 4x slower with the new clients, 
but if you built a new server with the new clients, then the old version was 4x 
slower than the new.  

Not to mention, I've had times when I couldn't even get linux & solars to 
*talk* to each other over NFS, due to version differences, nevermind tweak all 
the little performance knobs.

So my advice is to first eliminate any question about version / implementation 
differences, and see where that takes you.

zfs-discuss mailing list

Reply via email to