VM cold migration: Can I throttle the disk transfer rate?

[richard6121]

Hi all,

When using VirtualCenter to initiate a cold migration from one storage medium to another, is there a way to throttle the disk transfer rate? I'm finding that the entire LUN slows to a crawl when moving a VM to another LUN on the same array.

[sbeaver]

There is no way to throttle in this situation that I know of. How big are your LUNS and how many VM's do you have per LUN? What kind of storage are you using?

Steve Beaver

VMware Communities User Moderator

====

Co-Author of "VMware ESX Essentials in the Virtual Data Center"

*Virtualization is a journey, not a project.*

[richard6121]

It's (cough)5400rpm SATA(cough). It's our shelf for VM's that don't do much of anything. Seven drives in RAID 5 carved into two LUN's of 891GB each. Yes, I know the LUNs are too big.

I'm willing to accept slow performance on this shelf, but when doing a VM move, I watch the SQLIO bunchmark numbers on an affected VM drop from 1800-2000 I/O's/sec to 3-5 I/O's/sec. For all practical purposes, they lose access to the disk.

Again, I know that near-line SATA makes for lousy primary storage, but this makes me think there's a problem somewhere.

[RParker]

> It's (cough)5400rpm SATA(cough).

7 spindles of 5400 SATA, with 1.7TB? WOW, its a wonder you can run VM's at all... Even if you throttle it's going to be slow.. That's why you want SAS instead of SATA, SATA drives are good for storage, not stress or performance.

[richard6121]

We have two more shelves of 10k and 15k SAS for the VM's that matter. Unfortunately we have quite a few other VM's with legacy systems and data that get used only occasionally and we just don't have the budget to allocate SAS to these VM's.

Fortunately, the SATA performs a lot better than anyone expected provided we're not migrating VM's onto the LUN. It's the only way that this shelf of disk has let us down so far.

Watching the I/O's drop from 2k to 500 would be understandable. Watching them drop from 2k to FIVE baffles me.

[Texiwill]

Hello,

All VMFS writes are already throttled by the vmkernel, however, you could not get slower than that. You may want to transfer over the network to another node, then use that node to write back to the disk at a 'slow' period.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

CIO Virtualization Blog: http://www.cio.com/blog/index/topic/168354

As well as the Virtualization Wiki at http://www.astroarch.com/wiki/index.php/Virtualization

[mike_laspina]

Something is not right about the drop in I/O.

Is it iSCSI attached?

I get better performance than that on a 3 disk SATA raid Z ZFS Solaris iSCSI storage server.

[richard6121]

It's 2gb FC-attached. All of our SAS LUN's and SATA LUN's are on the same EMC CX500. I did the same kind of VM migration to a SAS LUN and saw very little performance hit. The drop was within the run-to-run variance of SQLIO (about 20% varianace because these LUN's are in production).

Again, I know that the SATA will underperform but I would not excpect a performance drop of 99% from doing a simple file copy on the VMFS.