Link to home
Start Free TrialLog in
Avatar of Thomas_Wray
Thomas_Wray

asked on

Should I present my storage to vmware vpshere 5 as one giant datastore?

Hello,
I am going to buy a large Dell powervault MD3200i device and i am wondering with the release of Vsphere 5 and its unified 1mb block size if there is any issues with presenting all of the storage to my ESX environment and formatting it as one giant VMFS datastore?
I know with our Netapp filer we were advised to keep 10-12 VMs per lun.

I wonder if the same holds true now I am going to use an MD3200i and Vsphere 5?

should i be doing the datastores in 2tb chunks or something?

thanks
ASKER CERTIFIED SOLUTION
Avatar of coolsport00
coolsport00
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
SOLUTION
Avatar of Andrew Hancock (VMware vExpert PRO / EE Fellow/British Beekeeper)
Andrew Hancock (VMware vExpert PRO / EE Fellow/British Beekeeper)
Flag of United Kingdom of Great Britain and Northern Ireland image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Avatar of Thomas_Wray
Thomas_Wray

ASKER

Thanks for your suggestions.
for the netapp route we typically went for one volume, one lun, one datastore per VM.

I am looking to get away from that as the wastage on our Netapp amounts to 2.4TB.

We will be buying 4x MD3200i boxes in an attempt to create resilience by having a number of boxes.
just like how you have multiple ESX hosts and can tolerate one failing we aim to have this but using MD3200i units.

I will be able to take one down if something goes wrong, using storage Vmotion.
Why do you think one large VMFS is all eggs in one basket? do they become corrupt?

unless somone has some horror stories or knows of a performance decrease with a large VMFS, then i dont see why not.

would love to hear more suggestions though :)

The chances of 1 LUN going down is remote due to the RAID configuration. But, it could. I always leans towards the side of caution. Also, if you have high I/O VMs on the same LUN, your performance could be degregated, unlike if they were on separate LUNs, and thus disks.

Again, there's nothing wrong with 1 large LUN, but you'll hear more of using somewhat smaller LUNs (datastores) than 1 large one. Recovery best practices is to not have all data, and in this case VMs, on just 1 datastore. But again, if you're comfortable with it, by all means, do so.

~coolsport00
Larger the LUN, the worse the performance, contention on the LUNs, with many VMs.

VMware Admins, that destroy the LUN, format the LUN, install VMware ESXi on the LUN (it happenes here on EE!)
as another point, i would be looking at potentially using the new features in vsphere 5 to monitor the response times on the datastores, this way if i know that one of my MD boxes is for highly responsive applications and has to have a response time of 10-15ms then vpshere will vmotion a VM to meet that rule.

this way i can have 4x identical boxes but Tier them automatically using Vsphere.

I am hoping i can get away with this! also i plan to monitor how loaded the MD boxes are using esxtop or just the response times on the datastores.

thanks
Yes, using Storage DRS.

Using information on storage capacity usage and (optionally) I/O response times, Storage DRS can shift virtual disks for VMs from datastore to datastore

please keep in mind that Storage DRS is an enterprise-plus feature in VMware.

Have you thought of buying EqualLogic Storage Systems instead of multiple MD3200's?

the MD-series is entry-level Storage - and your "4x MD3200" does not look like an entry-level scenario.


A very nice thing with multiple EqualLogic Storages:  they do load-balancing internally, so you do not have to use EnterprisePlus-Licenses for StorageDRS


And regarding your original question: planning of single/multiple LUNs should also always depend on your plan to use storage-level snapshots (storage-level snapshots are allways taken for a complete LUN, so if you are planning to use different storage-snapshot-schedules or settings, please seperate them in different LUNs)

I would consider other vendors before going the Dell MD series route..interestingly you are buying a Netapp owned product as the MD line was a rebranded LSI offshoot that got bought by Netapp.

Using one lun/datastore per VM on the Netapp is a waste of time, why didn't you put multiple VMs in a single datastore, whether it be NFS or block based?  The SIS feature typically reduces space usage by 40-60 percent on your typical utility VMs.  Unless you have very large VMs the architect of this design was not providing due diligence.
Originally the storage provisioning was suggested to us by the people who installed it 2-3 years ago, we were new to netapp then and VMware so it was considered the easiest way to break us in. now we know better but we are getting rid of our Netapp kit so we are looking for  a direction to go in, i thought the MD3200i would be able to perform happily but perhaps i am wrong?
I agree an Equalogic SAN would be great i am actually waiting on a quote/suggestion from Dell as they have our Vmware stats.
No i wouldnt be using storage level snapshots, we use Veeam to backup all our VMs
It was a very commom approach in the UK, as far back as 2004, to create single LUNs for single VMs, as very few Resellers were coming online in the UK, to Resell NetApp Kit, and "NetApp has just got into bed with VMware". Also if organisation wanted to restore a single VM, by using Snapshots it made it easier.