VA7400 performance issues

[Tim Crook]

Hi all,

Just want to see if anyone out there is having similar issues with the VA7400 as we are currently suffering...
We purchased a VA7400 a while back, been generally ok - had critical issues with it's original firmware (HP11) which has been resolved with HP14, but we're having some serious issues with performance; specifically I/O.
During operations we see I/O bottlenecks on such a regular basis it's become normal! - glance is coming back with warning/critical alarms with -alarming - (sorry) regularity and it hits 100% once or twice an hour for 10 minutes at a time. The VA7400 is attached via FC to a brocade switch, which, in turn in connected to two L2000's and an N4000. The N is the DB server (running Oracle) and is where all the I/O is coming from (90% of the VG's on the LUN's are on this server). With the help of HP we have got the VA configured to be as quick as possible, however, looking at all the stats, only one controller on the VA is taking all the hits (even though both have been configured to be used. We are going to try and load the redundancy groups more evenly by spreading the LUN's that are accessed frequently across both RG's therefore using both controllers more evenly. We've also been advised to install Auto Path onto the N and L's to help LVM cope with the array and to force both controllers on the VA to work together.
Is anyone out there suffering the same problems, have you managed to get around them by loading the RG's evenly and installed Auto Path? or are there any other methods...
That's as breifly as I can put it otherwise I'd be writing a best seller.
Thanks in advance!,

Tim

[Paul Barmettler · Accepted solution]

Hi Tim,

we had the same problem and made good experiences with putting the datafiles in RG1 and the logfiles into RG2.
So a lot of I/O will be shared over the two contollers.

Regards

[Tim Crook]

Hi Paul,

Good point - we'll keep this in mind when we start moving the LUN's across the RG's.
From your response it's good to hear that your average I/O throughput improved - was it considerable?

[Paul Barmettler]

Hi Tim,

yes it was considerable.

Here some other hints:

- set Data Resiliency on your VA to High Performance mode
- note that you configure the correct I/O path.
For Example: If your datafiles are lying on RG1 and your logfiles on RG2 ensure that the primary I/O path for the datafiles is configured to controller 1 and for the logfiles configured to controller 2.
- we made also good expiriences with changing the queue_depth.
Formula: Number of disk in RG * 2 / Number of LUNs in RG

P.S.: I've heard that the next firmware (HP15) will also more improve the performance of the VA.

Good luck

[Tom Danzig]

Tim,

Dump the thing. It's no good!

J/K ;) We'll straighten it out.

[Tim Crook]

Hi Tom,

Thought I'd saw you around here - you obviously come here a lot judging by your status!
Hopefully this last lot of work in the VA will be the last...touch wood.

Tim

[Vincent Fleming]

A new version of VA firmware is now available... HP15. It's supposed to help with performance (among other things). It's recommended that this be installed by an HP CE, which should be free of charge.

I just thought everyone would like to know!