EVA Performances

[Benoît]

Hi,

Does anybody have performances charts on EVAs?
What is the max I/O EVAs can perform? (just theoretical is enough)

Thanks

[Luk Vandenbussche]

Benoit,

It is in the quickspecs

EVA3000 :

http://h18006.www1.hp.com/products/quickspecs/11619_div/11619_div.pdf

EVA4000

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

EVA5000

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

EVA6000

http://h18006.www1.hp.com/products/quickspecs/12235_div/12235_div.pdf

EVA8000

http://h18006.www1.hp.com/products/quickspecs/12236_div/12236_div.pdf

[Benoît]

Good! Thanks!
I didn't see it before.

Is it possible to calculate in some way, depending of number of disks?

[Benoît]

Don't you find strange that IOPS are the same for EVA3000, 4000, 5000, and 6000?

[Thomas J. Harrold]

could be that the stats are based on "controller pair". I've seen different charts that show IOPS in various configurations. (ie number of shelves, disks, disk speeds, etc.

If I can find it, I'll re-post.

-tjh

[Vincent Fleming]

The 141k IOPS stated in the Quickspecs are maximums with 100% cache hits (all data is in cache). It's a measurement of processor speed and cache bandwidth and latency.

The actual number of IOPS you will see depends heavily on your particular workload and number and type of drives used.

Typical numbers you will see in the field are approximately 100-120 IOPS per 10K RPM drive, and 160-185 IOPS per 15K RPM drive, given a generic workload of 4-8k blocks and limited randomness.

So, if you work it out... 240 15k drives is about 42,000 IOPS at 175 IOPS per drive and no cache hits.

You may see better or worse performance depending on your individual workload - the cache can increase your performance exponentially if you get a good cache hit ratio.

Good luck,

Vince