autoraid 12h write working set

[Joseph C. Denman]

Hi all,

I've been looking that the admin and user guides for the 12h and have read a lot of threads. All of the info on performace states that the write working set value should be <= 1. I have set up my autoraid as king clay has advised on several occassion. Now I would like to check performance. I am use to using jamaca boxes and this raid sees slow. My question is how do I get this value? Is there a command or calculation I can use? Any help appreciated!!!

...jcd...

[Andreas Voss · Accepted solution]

Hi,

arraydsp does it for you:

arraydsp -m MMDDhhmm MMDDhhmm

Example for one day:

arraydsp -m 09180000 09190000

See also 'man arraydsp'

Regards

[Bill McNAMARA_1]

http://www.hp.com/cposupport/information_storage/support_doc/lpg75003.html

From man arraydsp:

arraydsp -m 04061200 04061800 2 00786b5c0000

PERFORMANCE METRICS
The following information describes the various performance metrics
returned by the -m option.

Metric Description

Drive Diff Values (Typical/Limit): 0-1/>1

Indicates if the disks are properly distributed
across the internal SCSI channels. If the disks are
not installed properly, one of the SCSI channels may
be handling more than its share of activity,
reducing performance. Redistributing the disks will
solve this problem.

Relocate Blocks Values (Typical/Limit): 0.008/>0.07

Indicates how much data is being moved between RAID
0/1 and RAID 5. If too much data is being transferred between RAID 0/1 and RAID 5, performance
declines.

Working Set Values (Typical/Limit): 1/>1.1

Indicates the ratio of the write working set to the
amount of RAID 0/1 space available. To maintain
performance, the amount of RAID 0/1 space should
equal or exceed the write working set (a value of 1
or less). A value greater than 1 indicates that the
working set is larger than the available RAID 0/1
space. In this case, the disk array must service
writes from RAID 5 space, which degrades
performance.

Cache Utilization Values (Typical/Limit): 0.1/0.2

Indicates how efficiently the write cache on the
disk array controller is being utilized. If the
write cache is consistently full, too many commands
begin stacking up in the cache queue, resulting in
decreased performance.

[Joseph C. Denman]

Thanks Andreas and Bill. As always, I always get more info than expected. You folks are the best.

Thanks again,

Joe...