Disk Alignment on guest VM's

[jrmunday]

Hi All,

We have a fair amount of legacy Windows 2003 servers with misaligned disks in our environment and I would like to resolve this to ensure that we are getting the best out of our storage, and not sufferring unnecessary performance degredation with something that can easily be avoided.

I did some testing to see how much of a difference IO alignment makes, and since the results are pretty significant, I would like to deal with this ASAP.

See results below;

So my questions are;

1. What is the best way to programatically go through all of the VM's and check for disk alignment - I don't want any manual process (if possible).
2. What are the options for aligning the misaligned disks?
3. What is the impact of each option (ie. How long does it take, Is Downtime required, etc.)
4. What are the limitations and / or known issues in doing such remediation tasks.

Thanks in advance,

Jon

[RParker]

meistermn wrote:

Why aren't there partitioning tools on the market ?

I do not understand the last part: all without rebooting the system.  Does this mean online realignment from hard disk to ssd. If so why not from vharddisk to vhardisk?

You are aligning the VM disk TO the SAN.  IT's a starting block, so they all have the same start / end points.  If you start at 100 and I start at 99 and our blocks are the same size (10 blocks lets say).

1 for block of data starting at 99 and one more for the 9 remaining blocks starting at 100. I will require 2 reads to your 1.  You are more efficient.  If I align my blocks to yours, we start / stop at the same point.  Thereby aligned and no overlap.

That's a very simplistic answer.. but that's how it works basically.  All we are doing is making everyone start at the same place... and since blocks of data is all the same (4K) cluster size, thus we can be in sync.

[RParker]

Matt wrote:

MOST array vendors (including NetApp and EMC) don't recommend defrag, because it moves data around the array uncessarily, with no benefit, and can even harm the automatic tiering algorithms.

Which is largely untrue.. we are NOT moving blocks around the SAN, we are simply defragging FILES *WITHIN* those same blocks... That's where everyone goes off the reservation... Defrag is at the GUEST OS level not SAN level... WAFL can still maintain the blocks ON the SAN, the files within the GUEST OS file system are *NOT* managed by the SAN, and therefore not part of the WAFL optimization..

The NTFS STILL needs to be defragged within a given block.  That's ALL we are doing.  Also automatic tiering is done over time (averaged).. so it's not going to affect anything.  Even if it does move SOME data.. ONLY the blocks that have higher IOPS get moved.. not ALL the blocks. so it's a very minor change.. in reality that's why we using the tiering method, get the higher performing IOPS OFF the lower performing spindles.. that's what's its for.  Once the IOPS return to normal those blocks will be moved back... once they are defragged, we don't need to KEEP defragging.  Start only when needed.

[jdptechnc]

I think EMC, NetApp, etc, do not come out in favor of recommending defrag to the general population because most lay-admins are just going to click-click-next and not consider the downstream ramifications of what they are doing (thinly provisioned disks being allocated more space unexpectely due to data being moved to previously unused blocks, blocks becoming hot and moving up a tier, or whatever).  It's easier for the vendor to just say not to do it unless you really know what you are doing.

[mcowger]

You may believe your theories.  They aren't reality, but enjoy them.

Nonetheless, neither EMC nor NetApp recommend it.    Is it possible they know more about the impact of defrag on their arrays than you do?  Just possible?

[cdnls]

Hi There

For anyone attempting this.

We used a combination of EBRALIGN, MBRALIGN and robocopy!

850 VMs (Linux and Windows) .

Change Management followed to the letter (god that was hard)
Started in June...finished in November!

The over time bill was massive!

[Gkeerthy]

for more information about the partitioning alignment and its internal please refer my post

http://pibytes.wordpress.com/2013/01/19/partition-alignment-in-vmware-vsphere-5-a-deepdrive-part-1-2...

http://pibytes.wordpress.com/2013/01/19/partition-alignment-in-vmware-vsphere-5-a-deepdrive-part-1/

[markzz]

There is a commercial tool available to complete disk alignments for you. The product was a written by vizioncore it's called vOptimizer.
This product is intended to perform many disk optimisation tasks in the virtual environment, one of these is start sector alignment, in fact it's the only feature I have ever used..
I have not used the product for over a year now.. Seems when vizioncore were purchased by Quest they went from being a friendly flexible company to a difficult, rigid and unresponsive one.
I've been reading this post in preparation for checking the alignment of our guest disks, we have completed a SAN migration and I feel the command latencies are higher than I'd expect..I therefore suspect during migration there may have been misalignment introduced into what was an optimised environment..
I'm dreading having to deal with Quest support..eg. I know we paid our maintenance on vOptimiser but as expected there was never a new license file sent for the new version.. I'm sure I can blow a couple of hours with support on this alone..
I'd hoped there was another product available which could perform disk alignments in a managed GUI interface..I've not found any yet beyond vOptimizer.

I see Quest has been bought by Dell.. hmmm they couldn't do any worse.

[JoelH]

Hello Matt,

I'm having a hard time getting copy of UberAlign. The download link for the OVA at nickopedia.com doesn't work. Do you know of a working download location?

Thank you,

Joel

[mcowger]

You'd have to ping Nick(@lynxbat on twitter) - it was his tool.  He doesn't work for EMC anymore (moved on to Intel)....

[TheVMinator]

Regarding NetApp, this document may clarify, though it is out of date an has not been updated since vSphere 4.x:

http://www.netapp.com/us/system/pdf-reader.aspx?m=tr-3747.pdf&cc=us

I would like to know if there is more current information or official documentation from them. Different technicians that come into NetApp start with as varied of backgrounds as those commenting in this post, and may have express opinions  differing from each other [disclosure - I used to be one back during the time this was current] , but before something gets published as this paper, it gets more of an official approval process.

Important I think in this document is this concept, that with VMFS there are two layers, not one layer of alignment:

1. The layer between the guest file system and VMFS, and

2. the layer between VMFS and the Array (not as much of an issue in newer VMFS versions)

Any thoughts on this paper?