05-25-2011 03:45 PM
I have a HP C7000 Chassis with two 4/24 Port Fiber Switches (i.e. Brocade 4024), two Xyratex SANs and many servers, VTL and Tape libraries. I'm trying to find out (1) what is the best way to perform I/O testing with the attempt of finding out what my queue depth should be and how much I/O the SAN will accept.
I'm also trying to find out if I have a throughput problem, when using the performance monitor that comes with the fiber switches, I find no more then 100-MB through put on just a few ports and no more then 1-GB aggregate throughput on the fiber switch. I'm wondering if I have a performance throughput problem and truly trying to find out how to test for it. I'm pushing data to 16 SAS and 16 SATA drives per SAN.
I just can't seem to find out what's wrong. I have placed calls to technical support and still no luck. So, I'm asking if someone can let me know their thoughts on what I could try next. Thanks for your assistance.
05-25-2011 09:39 PM
Do you think I should be seeing more then 10 or 100 MB/s on my SAN switch? I feel that there are bottle necks on the SAS and SATA disks.
If you ever get time, I would like to disucss this with someone that knows more about fiber channel and sans then I. HP support was not great help.
05-25-2011 09:42 PM
I have been testing out SQLIO and I have started to use IOMeter. Just not quite sure how these tools work. I'm hoping that these tools when I run to stress test the SAN, that I will be able to see a increase in I/O or throughput in the Brocade performance monitoring tool. That would make me feel good. Because at this time, I feel that there is a problem with both SAN switches or the HP C7000 back plane issues.
05-25-2011 11:56 PM
If your storage array is not able to handle the IOS from your servers the SAN switch will never shows higher through put.
Performance is relativ and depends most on your storage array and its config.
You have to think about RAID level and RAID group size this will affect performance. You have to know what kind off IO pattern your application produce.
Keep in mind that the rotation speed of your disks will have an massiv impact on responce times. This is important when you have random IOs,
It is difficult to say how much data you can push or pull from your box.
To make sure that you SAN switch is OK check on the CLI the error counter of your ports with porterrshow
before you run this command delete the error counter with statsclear.After the cleanup wait 5 to 10 minutes and run porterrshow.
If the Error counter are atzero level your SAN switch and the connections to the server and storage is OK.