VMware Cloud Community
zenomorph
Contributor
Contributor
Jump to solution

EMC Clariion CX3-80

Hi,

Were going to put our vSphere onto EMC CX3-80 with 600GB 15K disks RAID5. Our plan is to have about 30 disks so 2 EMC DAE.

What were not sure is the best way to configure the LUNs and RAID for optimum performanceeg. use metaLUNs or just across the same DAE?

Also what size and the VMFS should be and whether it would be a good idea to use VMFS extents.Typicaly we will have VMs where the OS drive is about 70GBs and the D:\ may range from 200GB to some larger 400-600GBs.

Many thanks

0 Kudos
1 Solution

Accepted Solutions
AndreTheGiant
Immortal
Immortal
Jump to solution

The reason I was thinking of using MetaLUNs across the 2 DAE's holding the 600GB disks to ensure we get can optimize more SPs rather having the LUNs across a single DAE using 1 SP.

This could be a good reason.

But you can reach quite the same by using more LUNs owen by different SP.

We'll probably be using RAID5 with 4-5disks per RAID group and a global spare.

If you have several disks consider also to build a RAID10 Raid Group.

But what about the VMFS block size - is there any impact on performance?

None.

But alignment could help. See:

In our case we'll most likely use the latter 800GB but does that also mean my VMFS should map to each LUN so 1VMFS : 1LUN.

IMHO 800 GB is a good size, and is better to have 1:1 for LUN/VMFS and Raid Group.

But should VMFS extents be used

I prefer to do not use extents.

And also if some VMs have d:\ e:\ f:\ that may total to larger than 800GB what would be the best approach

Consider to use RDM disk in this case.

Andre

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro

View solution in original post

0 Kudos
4 Replies
AndreTheGiant
Immortal
Immortal
Jump to solution

I suggest to use different RAID gruop and (if possible) build a single LUN on each RAID group.

Do not use meta-LUN.

Use at least a global host spare (if you have different disk speed is better to use one GHS for each speed).

On EMC powerlink site you can find several best practices.

See also:

Andre

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro
zenomorph
Contributor
Contributor
Jump to solution

Andre,

The reason I was thinking of using MetaLUNs across the 2 DAE's holding the 600GB disks to ensure we get can optimize more SPs rather having the LUNs across a single DAE using 1 SP.

We'll probably be using RAID5 with 4-5disks per RAID group and a global spare. But what about the VMFS block size - is there any impact on performance? And I read the doc link on recommending LUNs of 400-800 in size. In our case we'll most likely use the latter 800GB but does that also mean my VMFS should map to each LUN so 1VMFS : 1LUN.

But should VMFS extents be used or recommended if I have some VMs that in case the might have d:\ drives larger than 800GB would using extents be recommeded. And also if some VMs have d:\ e:\ f:\ that may total to larger than 800GB what would be the best approach and should I have seperate LUNs/VMFS for System and Data?

Many thanks

Much appreaciate your advise.

0 Kudos
AndreTheGiant
Immortal
Immortal
Jump to solution

The reason I was thinking of using MetaLUNs across the 2 DAE's holding the 600GB disks to ensure we get can optimize more SPs rather having the LUNs across a single DAE using 1 SP.

This could be a good reason.

But you can reach quite the same by using more LUNs owen by different SP.

We'll probably be using RAID5 with 4-5disks per RAID group and a global spare.

If you have several disks consider also to build a RAID10 Raid Group.

But what about the VMFS block size - is there any impact on performance?

None.

But alignment could help. See:

In our case we'll most likely use the latter 800GB but does that also mean my VMFS should map to each LUN so 1VMFS : 1LUN.

IMHO 800 GB is a good size, and is better to have 1:1 for LUN/VMFS and Raid Group.

But should VMFS extents be used

I prefer to do not use extents.

And also if some VMs have d:\ e:\ f:\ that may total to larger than 800GB what would be the best approach

Consider to use RDM disk in this case.

Andre

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro
0 Kudos
chilow
Enthusiast
Enthusiast
Jump to solution

If you end up using RAID1/0 make sure you seperate the data and mirror drives between DAEs.

-C

0 Kudos