Volume Sizing Performance

[LightSpeedHost]

We have a 8540 all flash we're implementing into a production environment as a first test of the 3PAR products that we'd like to shift too.

In the case of our demo we have a Xen cloud cluster (one pool for AMD CPU compute notes and another pool for Intel).

Usually what we'd do in ZFS was setup a pool for each compute pool.   So we would have 2 pools that are about 30TB in side for each pool.

For this demo, if we create a large volme, say 20TB (iSCSI) and then format it and use it this would fit our normal operations.  However, my reading has indicated that 20TB would be a HUGE single volume and I am wondering if there would be a huge performance hit by creating such a large volume?

If there is a huge performance hit, what is the suggested utiliztion when working as block storage for VMs?  I can't imagine we want to go through the process of creating a volume in the 3PAR to mimic each VDI we create in Xen?

Thanks!

[sbhat09]

Hello @LightSpeedHost,

3PAR supports volume size up to 64TiB. So, 20-30 TiB is fine from the compatibility/support perspective.

If you are doing a high IOps performance testing, please ensure enough ports and bandwidth are available to handle the high amount of read/write operations from several VMs together. So that there is no bottleneck at ports level.

Regards,
Srinivas Bhat

If you feel this was helpful please click the KUDOS! thumb below!
Note: All of my comments are my own and are not any official representation of HPE.

[LightSpeedHost]

Thank you.  I am putting this into xen cluster today and running some testing.  I've been using a dedicated dual 10Gbps server to throw IO load on it and today I am hoping to spin up a bunch of VM's against it, use MPIO and see if I can max out the four 10G links on this unit and see how it performs.  It's HIGHLY unrealistic that a real world scenario would ever look like this but I am super curious to see what it can do ha!