Monday, September 29, 2014

Nimble Storage Upgrade with VMware Horizon View 6 Rebalance


Hi Friends,


Today I'd like to share a really cool process I performed the other day.  I was lucky enough to get some new storage arrays and I'm moving my Horizon View testing from one array to another so I can run performance tests with Login VSI using the new Nimble CS300 array.

I decided not to use storage vmotion, but instead use the rebalancing tool built into Horizon View 6 to transfer from one array to another.  The cool thing is you can use this procedure not only to transfer from one array, but increase space to existing datastores, add new datastores to View, etc.  The Rebalance feature evenly redistributes linked-clone desktops among the datastores that have been assigned to the storage pool.

What I thought I would do is show you how easy it is to create a new volume on Nimble, assign it to my vCenter 5.5 environment and than show the Rebalance process in View 6.

1.  This is the Volumes page in the Nimble GUI.  I've already created one 2TB volume called cs3-vdi1.  I'm now going to show you how to create another volume called cs3-vdi2.



2.  The first step in creating a new volume is giving it a name and then selecting a Performance Policy.  This is a super cool feature in Nimble.  Say you have an application that writes in 8K blocks, or 32K, or 16K.  You can customize how your Nimble will write to the volume!  Nimble provides a bunch of default policies of popular apps, but if you have something that's not in there, you can create your own.



3.  Let's select a size for our volume and below that is settings for customizing volume and snapshot behaviors.  What you see below are the defaults, but you can change these settings depending how you want to use thin provisioning, alerts, snapshots, etc.



4.  Now we select how we'll protect our new volume.  Here I'm choosing not to protect it since my linked clones will live on this volume.  It's a good idea to create snapshots of your infrastructure, user data and your golden image.



5. We're all done and you can see the new volume has been created.  Pretty easy eh?  :-)



6.  I won't bore you with adding the volumes as datastores to vCenter, unless you want me to.  Here both of my new volumes have been added to vCenter as datastores, ready to use!



7.  I'm in Horizon View now, I've gone to my Desktop Pools, selected my VDI pool and clicked on Edit and will browse for the new datastores on my new CS-300.



8.  Here's where we have a bunch of choices.  You can see the two new datastores from my CS-300, and the two datastores from my CS-460.  I can just add the two new datastores to my pool and get an additional 4TB to work with or move my desktops onto my CS-300.




9.  In this blog I'm showing you how to rebalance onto a new array, so I'm going to deselect my CS-460 datastores and only select my new CS-300 datastores.  I could have also re-sized my datastores, or chosen a bunch of different combinations depending on what my storage needs are.



10.  After I've decided how I want to use my storage, I next select the View Composer Rebalance.  Since I've decided to use two new datastores and not the original two, the desktops will be moved from my CS-460 and be evenly redistributed across my two new datastores on my CS-300.



11. And away we go!!



12.  Let's take a look at the two storage arrays and see what's happening during a Rebalance.

13.  Here's a graph from my CS-460.  You can see it's performing a lot of small block sequential reads and random writes.  Our guess is there is a lot of coalescing and cleanup going on during the Rebalance.

 


14.  This is a graph from the CS-300.  The interesting thing is there is a lot of both sequential and random reads going on.  Looks a lot like a boot storm huh?  Our guess is there is a lot of checksums and validation going on here to make sure all of the data has been properly moved.  The interesting thing is there is only a small amount of writes...  This is probably by design if you think of the way Linked Clones work.  They are de-duped by design by using only a very small delta disk and referring back to the golden clone for the operating system.  This way they are very storage efficient only using a small amount of space.



I hope you enjoyed the blog as much as I enjoyed testing it!

Until Next Time,
-Brain