> From: David Weibel
> Sent: Tuesday, June 19, 2007 9:49 AM
> 
> 1)  Are there Errors.  Checked /var/adm/messages for any type 
> of errors during your IO run.  Check netstat for network errors
> and collisions.

No errors and the network is idle except for this traffic (test lab for storage 
exclusively.)


> 2)  Simplify the Problem.  Take ZFS out of the picture.  I'm 

Since posting the other day, I have done this.  Performance on the network is 
broken in some way, even after setting jumbo frames.  I am investigating this 
now.

Now that the 10GbE and switch are configured with jumbo 
frames (9198 on switch and 9000 on T2000/X4500 systems) I am 
still experiencing slow performance across the network.
- iscsi targets capable of 350+ MB/s local on X4500 give 20 MB/s T2000->X4500
- java TTCP on 10GbE -> 110 MB/s and 4x1GbE -> 90 MB/s
  (10GbE -> 16000 p/s and 4x1GbE -> 10000 p/s with both boxes < 5% utilized)

Apparently the problem is (mostly) on the t2000... the numbers for X4500->X4500 
are better. (I say mostly because it a little more than half what I would 
expected on 10GbE/PCI-X.)
- java TTCP on 10GbE -> 325 MB/s and 4x1GbE -> 95 MB/s

I appreciate the pointers to the iscsi/network tuning items and will work to 
incorporate them into the testing.

Will report back when I resolve the network problem and apply this information.


--

paul
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss

Reply via email to