Thursday, October 25, 2012

Why Enterprise Storage for Writecache?

Hi Everyone,

I'm often asked, "Why should I put my writecache on your storage?  It's only transient data."  While I was somewhere over the Atlantic I had a thought (yes, yes, occasionally I do get them.) regarding RPO and RTO.  For those of you who don't know, these are terms used in disaster recover and backup recovery.  RPO is the Recovery Point Objective and RTO is Recovery Time Objective.  Basically, how fast and at what point in your data do you want to recover from.  So what does any of this have to do with the writecache?  Good question!

Our best practice states you should put the writecache on disk so you don't overwhelm your PVS server and for lots of other good reasons.  Yeah, yeah, I know there will be tons of people saying, "Just put the writecache on the server." but that just raises a whole new set of problems.  Even though that disk is just transient data, it's still an integral part of the OS, so I asked my performance engineer to try a test!  He had 100 desktops running LoginVSI and he pulled the writecache datastore.  Things were not pretty after that.  The desktops hung and were even grayed out in vCenter.  vCenter still functioned, but it was slow and clearly did not like we had just done.

So imagine your writecache is on a less expensive storage that has no fault tolerance.  Will user data be lost?  Probably not, but your users will be stuck until you can get something online again.  Hence RTO.  We're always so concerened about protecting data, that I think we forget about protecting the infrastructure too.  Think of this as the public transit of VDI.  Writecache won't make you the new super start up, but it will get you to work so you can focus on becoming that super start up.

Hey, for more information on NetApp and PVS take a look here:
NetApp and PVS

Its a great paper written by Rachel Zhu and myself.

No comments:

Post a Comment