Disk Arrays
cancel
Showing results for 
Search instead for 
Did you mean: 

AutoRaid 12H - normal behaviour

AutoRaid 12H - normal behaviour

We are running an AutoRaid 12H attached to an L class running HP-UX 11.

It it fully loaded 2 Controllers, 6 x 18Gb and 4 X 4Gb disks. There is currently ~30Gb un allocated and a hot-spare configured.

We notice that if two processes are making I/O requests the performance of each is impacted. e.g. Oracle RDBMS doing writes with fbackup doing reads at the same time.

My question is

Does AutoRaid typically stripe all lvols across all phyiscal disks, or is that only when in RAID 5 Config.

Is there a way to check which RAID mode is in operation at a point in time.

Is the performance degradation normal/expected when there is two or more I/O operations happening simultaneously.

Is there any way of controlling the placing of lvols on physical disks, DBA's like to ring-fence data from indexes etc.

regards

MJW
4 REPLIES
Xavier Gutierrez
Frequent Advisor

Re: AutoRaid 12H - normal behaviour

Hi, MJW.

Does AutoRaid typically stripe all lvols across all phyiscal disks, or is that only when in RAID 5 Config.

Yes, *IT ALWAYS DOES*

Is there a way to check which RAID mode is in operation at a point in time.

No, there is always data on RAID-0/1 mode and on RAID-5 mode. You can only see how much data is each RAID level (amdsp)

Is the performance degradation normal/expected when there is two or more I/O operations happening simultaneously.

Don't think so, but can't explain further.


Is there any way of controlling the placing of lvols on physical disks, DBA's like to ring-fence data from indexes etc.

No, there is no way.

Regards, XGM
Live fast, die young!
florence mathon lermusi
Trusted Contributor

Re: AutoRaid 12H - normal behaviour

Hi,

the only thing i know about this type of array is that you can have a degradation of perf if your % of disk space allocation is greater than a certain % of the total space available because the array is automatically reconfiguring the raid level used... In this case the way to get your optimal perf back is to add some disk in the array or to put some larger disk in it...
A. Clay Stephenson
Acclaimed Contributor

Re: AutoRaid 12H - normal behaviour

The fundamental problem is that you have allocated too much space as LUN's. These old boxes perform far better when then total allocated space does not exceed about 50%. The AutoRAID will then remain in RAID 1/0 at all times. The other key to performance is to utilize both external SCSI paths. Ideally, you divide each VG into 2 LUNS of equal size. LUN A's primary path should be through Controller X (alternate Y) and LUN B's primary path should utilize Controller Y (alternate X). You then stripe each LVOL in the VG across these two LUNS typically using 64K blocks.

If you follow these procedures, you will find that the old 12H's actually perform quite well even under moderately heavy loads.


You have absolutely no control over have each LUN is allocated across disks and indeed this can change dynamically.
If it ain't broke, I can fix that.

Re: AutoRaid 12H - normal behaviour

While doing unrelated maintenance to-day we noticed that the server has a dual-port scsi card, one port has one of the controllers for the Array connected and the other has the external DLT tape drive attached.

as (bad) luck would have it the primary access path for most LUNs is via the controller sharing the SCSI card with the tape drive.

Would the card be able to cope with simultaneous heavy disk reads and tape writes as happens during nightly backups sometimes when they overlap the batch-run window. Or is the SCSI card unlikely to be a bottle-neck