As we are preparing for the GA release of XtremIO, we already have many production customers, one of the no brainer use cases is using XtremIO for VDI and i see a lot of demand for XenDesktop 7.

while im working on large scale White Paper (Stay tuned!), i wanted to draft some of the considerations you want to think about when using XtremIO with XD7 and Provisioning services (PVS)

1. When using MCS / PVS, please set the “Hosting” parameters to accommodate these numbers (you can actually increase them, depends on the customer Compute resources that are available)

2. When using PVS, make sure the vDisk is aligned to dynamic 16MB which are 4kb aligned as shown below

3. PVS have multiple ways for the write cache, when using a SAN based environment, you will probably going to use the “Cache on device hard drive” in order to increase performance, otherwise, selecting “cache on server” will not result in a good performance

In order to do so, you will have to take the VM that you are going to to convert to a template, yes, it’s the one you installed windows on and then used the XenTools to convert to the master VM.

4 Comments »

  1. Hi Itzik,

    I have a client that is using ExtremeIO and plans to use a 60GB PVS write cache. On boot of , they will perform a sizable copy of the App-V sequenced apps, up to 20GB. They plan to use dedup. There will be 12 XenApp VMs per host and 6 hosts in the immediate term. I think the approach for copying the sequenced apps on boot is a bad idea but that’s a separate discussion. My question is can ExtremeIO handle this configuration if all servers are rebooted at the same time?

    Thanks in advance!

    George

    • Hi George,
      i think that there is some confusion here since I’m only familiar with a product named XtremIO 🙂
      seriously though, i don’t see a major issue with the approach, from a boot perspective, the number of VM’s (12 X6=72) shouldn’t take more than a few minutes! the application copying will probably take longer but still, it shouldn’t be a major issue because of the TRUE inline dedupe, the latency will probably be higher during the copy process as it’s normally using very large block size.

      hope that’s answer your question
      Itzik

Leave a Reply Cancel reply