Cloning Very Slow With Remote ESX Server

[fitzie22]

I have one virtual center 2.0.1 patch 2 in boston and 10 esx 3.x servers in it also in boston. If I clone a vm from one server to another it works fine (10 min). I then have one esx 3.x server in Maryland connected to the VC in boston and if I try and clone a server on that esx server to its self it takes an hour. the network is at a gig in both location and a 45meg pipe between the sites so I am pretty sure there is not a network latency problem. any ideas where to begin?

Message was edited by:

fitzie22

[davidbarclay]

Slow cloning and template deployment usually points to disk i/o issues. Anything unique about the remote disk config?

Dave

[fitzie22]

nothing unusual all the servers are the same make and model with one large RAID-5 array

[gogogo5]

is your one ESX server in Maryland using local storage? If so, I wonder if the performance is hampered by the I/O of the clone going to/from the same storage.

[fitzie22]

yes it is local storage so you are correct the the cloned image is going from and to the same disk

[gogogo5]

maybe a clue then, not conclusive but probably attributable...

how large is the VM you are cloning?

[fitzie22]

2 disk, 8gig and 4 gig

[gogogo5]

ok, so not an extreme size. If you want to put your mind at rest and you have a suitable server or time window you could mirror your clone task on one of your ESX hosts in Boston to/from local VMFS. Then you will have more of a comparison, especially as you mentioned all your hosts are the same spec.

[fitzie22]

Ohh I already have and it only takes about 15 min or so. So that is why I was wondering if it had to do with th VC being in boston.

[gogogo5]

But in Boston are you cloning to/from local VMFS on your ESX hosts too or are you using a shared storage or SAN architecture and performing the clone on that?

[fitzie22]

ok So it looks like that is what the issue was. cloning the same size disk on local disk takes between 45-55 min and the same cloning on the san takes 11 min. WOW is that a difference. This will probably help the argument I have to move to the SAN. Thanks for all your help. actually do you have an explenation as to why that is the case?

[gogogo5]

the local VMFS copy on the Maryland server will be going through the same controller and disks and therefore read and write operations will be contending for I/O. Also, the RAID level you use can definitely affect performance, especially on combined read/write operations.

Without knowing too much about your SAN you could be using superior disks, spindle speeds, and potentially different LUNs for performing the clone. All this will contribute to faster clones.

BTW - what SAN do you use?

If you have found the help useful in this post please award points, thanks!

[fitzie22]

not 100% sure but the san is a clarion 200 someting

[davidbarclay]

Cloning involves creating and then growing a VMDK. If the VMDK is on the same LUN (the one and only LUN), each bit of growth causes a LUN lock...this would definately slow it down.

Dave