Disk Arrays
cancel
Showing results for 
Search instead for 
Did you mean: 

Best Practise (HP EVA - 6400, HSV400 ) for Vshpere 4.0

virtualtechie01
Occasional Visitor

Best Practise (HP EVA - 6400, HSV400 ) for Vshpere 4.0

All, storage experts.

 

Would some of you would be able to shed some light on below.

 

With EVA's it uses disk virtualisation aggergates right, for the Vshpere environment what would be best practise which would have a combination of SQL, windows, web and other I/O intensive servers.

 

Basically we have 2 disk groups, 1 for FATA and one for FCAL --- we have  86 - FCAL disks (558GB) where all the VM's reside. most of which are Vraid5, there are approx 52 Vdisks (LUNS), vraid5 mostly, and we have around 110 virtual servers (probablly about 150-200 vmdk's) which are all sorts sql, web, windows, backups

 

So the question is - is one disk group FCAL with 86 disks is good enough? from the perfomance i cant see it being good enough but whar are you guys reccomending storage wise?

 

Any tools for EVA so that I can monitor if the disk are being Thrased?

Any sizing or best practise guide for Vshpere4 (combo of all abov servers types)

or would it be best to create differnt disk groups for diff server types (SQL, web etc)??

 

Please advise experts..

Cheers

2 REPLIES
Johan Guldmyr
Honored Contributor

Re: Best Practise (HP EVA - 6400, HSV400 ) for Vshpere 4.0

Hey, 

 

evaperf can give you lots of information. It comes with the Command View installation. HP support has a nice tool to make graphs etc.

 

http://h20195.www2.hp.com/v2/GetPDF.aspx/4AA1-2185ENW.pdf

 

Looks like a good document, it's called "HP ENTERPRISE VIRTUAL ARRAY FAMILY WITH VMWARE VSPHERE 4.0 , 4.1 AND 5.0 CONFIGURATION BEST PRACTICES"

 

There are more white papers here: http://www.compaq.com/storage/whitepapers.html

 

You can have max 16 disk groups in an array.

 

" from the perfomance i cant see it being good enough but whar are you guys reccomending storage wise?"

Do you mean capacity wise? As in best way to utilize the storage space?

 

If you have a server that could (for example) at peak utilize all the available disks' IOPS (or a large chunk of it), then I'd start thinking of ways to isolate that application from affecting the other servers.


But if all your services can co-exist on this one disk group, why not? Isn't one of the greater benefits with ESX that with this shared disk you could move the VM to another physical host?

Steven Clementi
Honored Contributor

Re: Best Practise (HP EVA - 6400, HSV400 ) for Vshpere 4.0

There are several white Papers on Vmware Best Practices when dealing with EVA Storage.

 

The short story is... many different environments have many different requirements, but there only a few choices.

 

For the best performance of a disk group on an EVA... you want to have as many drives in the group as possible.  More drives = more iops availability to service requests.  Best Practice here is to have only 1 Disk group.  Obviously you can not group FATA disks with Fibre Disks so the 2 disk groups are warranted.

 

vRaid5 vs. vRaid1... a vRaid5 disk uses @ double the amount of iops to write data to the group.  vRaid1 obviously will use up more space, but performance would be better in a tight iops situation.

 

How are your datastores laid out?  How many Vm's do you have any any one particular datastore?  Are you grouping Database VM's? or separating them?

 

There are lots of little tweaks and such to make in your environment to allow it to perform better.

 

As mentioned... EVA Perf comes with Command View.  Collecting data is relatively easy, reading that data might prove to be a challenge if you've never done it before.

 

Best Practices and such White Papers:

 

http://www.vmware.com/files/pdf/techpaper/hp-enterprise-virtual-array-family-vsphere-configuration.pdf "

http://h20195.www2.hp.com/v2/GetPDF.aspx/4AA2-8848ENW.pdf "

 

The second paper might have some useful information in it for optimizing performance.

 

 

 

 

 

 

Steven Clementi
HP Master ASE, Storage and Clustering
MCSE (NT 4.0, W2K, W2K3)
VCP (ESX2, Vi3, vSphere4, vSphere5)
RHCE
NPP3 (Nutanix Platform Professional)