Perfomance baseline for XP12000

[Keith C. Patterson]

Hello all,
I was wondering if anyone can provide a standard methodology/strategy for baselining performance on a new XP12000? Everything is configured with RAID 5. Some LUSE volumes also.
Since it would be next to impossible to provide numbers prior to going into production I would like to provide some recommendations on the key metrics to monitor from the XP side and the host side. Note, this is not application specific, i.e. Exchange, Oracle.
I would assume metrics such as read hit ratio, IOPS, transfer rates etc would be a good place to start. If you have any numbers that would be even better.
Also, is there a rule of thumb for how long one might recommend performing baselining in order to get some solid numbers?
I realize that this may be a loaded question but I was hoping that I wouldn't need to reinvent the wheel.

Thanks in advance.

[sivam]

XP12000 Perfomance is based on the I/O distributed across the CHIP Ports. Each CHIP Ports can have a maxmimum of 1 Gig of Data Transfer for better perfomance. As specifically you can limit the I/o to 750MB per Chip Port.
XP Arrays are based on Cache Memory Read & Writes so the perfomance is always there unless untill the I/o is heavily loaded.

If you have any more please let me know.

sivam

[Peter Mattei]

Have a look at the HP StorageWorks XP12000 Disk Array best practices for performance guidelines white paper on:
http://h71028.www7.hp.com/ERC/downloads/4AA0-7923ENW.pdf

It will help you a lot!

I strongly recommend purchasing Performance Advisor XP. It will enable you doing performance monitoring (real time and historically) of every part of the XP and help running an XP for years and keep performance up by detecting bottlenecks upfront!
See http://www.hp.com/products1/storage/products/disk_arrays/xpstoragesw/performance/index.html

Cheers
XP-Pete

[avik]

LUSE means concatenation, not striping, if the LUNs were striped from the host end, you will see better performance out of a LUSE volume spread across array groups versus contained in a single array group.