HPE Storage Users Group

We just got our 7200 and I will have my course next week. It will be installed by an HP partner.
At the moment 60 VMs are running on the old storage (IBM DS3512, 2x 45TB Datastore) but the storage cannot cope with the IO and runs out of guarantee also. The 7200 will be it's replacement and exclusively used for 3 ESXi5.5 Servers to host about 75 VMs.

Config of the 7200
64 900GB SAS 10.000
36 3TB Nearline SAS 7.200
12 1.92TB cMLC SSD
UPDATE: WRONG, not included -> 2x4 1,92TB cMLC SSD
I did not do the config or buying, I am just the guy who has to work with it ;)

So maybe you have any tips&ticks for me regarding the 7200 on one hand and combined with VMWare 5.5 on the other hand. What we plan is one big datastore and let the 7200 do it's magic most of it's own (at least this was like it was presented).

Thank you for your time :)
Joachim

Have one question.
What summ you pay for this?

Georgii wrote:Have one question.
What summ you pay for this?

I work at a non-profit government-funded research center, so we get quite special prices. I am not allowed to disclose them though, sorry.

That is one powerfull little array... keeping it "kiss" is most important. usually the more you tinker with it, the more you break.

HPs ordering system is a bit ugly, so sometimes drives are broken up over multiple ordering lines. i've seen that before. I assume you "just" have 20 2TB drives.

On the host side: just present each node via 1 path to the host. 4 paths would not make much sense with a 2-node array. (assuming 8gbit host ports)

i'd start simple. configure 3 cpgs for each of the tiers, delete the default ones you do not need.
reserve space to use AFC, but don't enable it just yet.
configure and enable AO
Migrate all data to the AO-FC tier, and let AO settle in. once AO is settled and all the hot data is on SSD, enable AFC.

Thing is, if you enable AFC from the start, it will offload a lot of IO, which prevents AO from moving it up to the proper tier. by first letting AO do its job you put the hot IO where it belongs, keeping AFC free for more IOs from the lower tiers.

for the rest, read up on: http://h20195.www2.hp.com/V2/GetDocumen ... c=us&lc=en

if you want to deviate from the best practices advices described in the document above, please post back here first. So we can understand why, and advice if that is a smart thing to do or not.

Architect wrote:On the host side: just present each node via 1 path to the host. 4 paths would not make much sense with a 2-node array. (assuming 8gbit host ports)

No, don't do that! Each node should be reachable via each fabric so you should have 4 paths. In fact you violate best practices if you don't and can interfere with Persistent Ports which can lead to data access problems during upgrades and node failures and with 3.1.3 port failures.

Thank you all for your responses. I will take all infos to my course and inform you then what I will probably do. Installation is in two weeks done by HP, I am quite excited

Architect wrote:Migrate all data to the AO-FC tier, and let AO settle in. once AO is settled and all the hot data is on SSD, enable AFC.

Thing is, if you enable AFC from the start, it will offload a lot of IO, which prevents AO from moving it up to the proper tier. by first letting AO do its job you put the hot IO where it belongs, keeping AFC free for more IOs from the lower tiers.

Do you have any advice on the time it needs to "settle in"?

afidel wrote:No, don't do that! Each node should be reachable via each fabric so you should have 4 paths. In fact you violate best practices if you don't and can interfere with Persistent Ports which can lead to data access problems during upgrades and node failures and with 3.1.3 port failures.

HUH? please reread the best practice guide! i think you are mixing up port persistence with zoning here. There is no need to zone all the port persistence members to a host at all. NPIV will take care of that, as long as you practice WWN-based zoning and not port based zoning.

Joachim wrote:Do you have any advice on the time it needs to "settle in"?

You can keep an eye on the amount of data moving to SSD. Either when SSD is full, or when roughly no more data is moved to SSD (the move-in and the move-out is about the same), the system is in balance. It depends on the amount of data and how busy the system is, when AO will reach this state. might be 3 runs, might take 10-15 runs.