Maximum total host KB/s for the EVA4000

[Timo Steylen]

Does anybody know what the maximum total host KB/s is for the EVA4000 or the maximum total host Req/s?
I have collected data with EVAPerf and the EVA shows an average of 7500 total host KB/s and a average total Req/s of 1000. The costumor is complaining about the performance.

thanks

[Víctor Cespón · Accepted solution]

http://h18006.www1.hp.com/products/quickspecs/12234_div/12234_div.html

Up to 141K IOPS and up to 335 MB/s throughput per Controller Pair

Of course that's the limit of the controllers CPU and RAM, the real limit is the type and number of disks.

On EVAperf you should look at the read and write latencies on the vdisk or disk group involved.

[Timo Steylen]

thanks!

The concerning vdisk has a much higher read/write latency in relation to the other vdisks.
What does it mean? An OS related problem?

[Víctor Cespón]

Several things have influence there. The OS has buffers for each LUN, the FCA in the server has also a buffer, and the amount of physical disks limits the number of IOPS the disk group can handle.
If this vdisk has the 90% of the traffic, it will have more latency than the others.

Check this page:

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c01671044

Multiply by the number of disks on your disk group, compara that to the number of I/Os on the disk group.

[Timo Steylen]

thanks again!

but what counter in EVAPerf for the diskgroup indicates the IOPS, the req/s?

[Timo Steylen]

this means if you use 56 10K disks, that the IOPS will be about 4648.

if i use the following formule:
MBâ sec = IOPs à ALLOCATION_UNIT_SIZE_IN_MB

4648 * allocation size (64K) that i have about 300 MB/s

Is this correct?

[Víctor Cespón]

That would be reading or writting to a VRAID 0. VRAID 1 and VRAID 5 can do less IOPS. But, generally speaking, yes.