Storage Boards Cleanup
To make it easier to find information about HPE Storage products and solutions, we are doing spring cleaning. This includes consolidation of some older boards, and a simpler structure that more accurately reflects how people use HPE Storage.
Disk Arrays
cancel
Showing results for 
Search instead for 
Did you mean: 

EVA 4100 - VMware vSphere 4- Max size vdisk LUN for without degrading performances

SOLVED
Go to solution
Mario Giaretta
Frequent Advisor

EVA 4100 - VMware vSphere 4- Max size vdisk LUN for without degrading performances

Hello,i know practical the best ones in order to configure the vdisk of EVA and I know that he convient to create many LUN of approximately 300gb assigning it to VMWARE. No document clears what is the maximum sizing that not cause degradation performance. First case: I create 3 vdisk about 300gb everyone and the check to vmware.
According to second case I create one vdisk only sizing 900gb and assigns and it to vmware. Do i have a degradation of performance in according to second case because to vmware must manage a volume sized 900gb? thanks Regards Mario
3 REPLIES
Rob Leadbeater
Honored Contributor

Re: EVA 4100 - VMware vSphere 4- Max size vdisk LUN for without degrading performances

Hi Mario,

Without knowing the details of the number of physical servers, number of virtual machines etc. it's probably difficult to say...

Cheers,

Rob
Uwe Zessin
Honored Contributor
Solution

Re: EVA 4100 - VMware vSphere 4- Max size vdisk LUN for without degrading performances

There is no simple rule of "large VMFS = bad performance".
A single VMFS partition can have 2 TeraBytes - 512 Bytes (=4,294,967,295 blocks of 512 Byte). And then you can 'link' up to 32 VMFS partitions (or extents) to one large datastore... (well, I think you need to tune the VMkernel a bit, but most people don't do that anyway ;-).

However, a VMFS (and all its extents) use a single I/O queue. So if you put many VMs which require lots of I/Os it is possible to run into a problem.

Another problem can be SCSI reservations: VMFS meta-data changes are serialized using SCSI reservations. Many events can cause that: VM power-on/off, snapshot creation/deletion, file creation / deletion / expansion(think about growths of DELTA files or thin-provisioning).

So large(r) VMFS = bad? Not necessarily. The other extreme would be to create a single VMFS per VM, but how do you size that VMFS if you want to use snapshots from time to time?
A good argument FOR large(r) VMFS is that it consolidates free space.
.
raj2796
Advisor

Re: EVA 4100 - VMware vSphere 4- Max size vdisk LUN for without degrading performances

Notes from when i went to vmware HQ in the uk and one of thier senior engineers gave us a great techy talk - this was when esx 3.5 was around though - might be different for vsphere

number of vms per lun
6 - 8 is the "sweet spot" for optimal performance
also past this point you being to encounter scsi reservation errors/locks

absolutely best performance - 1 vm per lun

personally i use 1 vm per lun for mission critical vms and 6 vms per lun for second tier systems

disk size
500gig is optimal - past this you see degredation

remember to use pvscsi nowadays for windows and linux - you'll see 12% more throughput and 18% less cpu use - i posted a load of articels on pvscsi at my blog :

http://raj2796.wordpress.com/2010/02/18/when-to-use-vmware-pvscsi-and-when-to-use-lsi-logic/