Virtualization Short Take #22

Despite the fact that I’m out of town this week at NetApp Insight, I wanted to go ahead and get out the latest installation of Virtualization Short Takes—my sometimes-weekly collection of interesting or useful links and tidbits.

  • Much ado has been made about VMware’s acquisition of Trango and the announcement of VMware MVP (Mobile Virtualization Platform). Rich Brambley has a great write-up, and I completely agree with Rich and Alex Barrett about what this really means: don’t expect to see Windows XP on your smartphone any time soon. Alex said it best: this is virtualization, not emulation, and Windows XP doesn’t run on ARM.
  • I’m curious—how many people agree with my comments in Alex’s article about the Citrix ICA client for the iPhone. Is there any real, actual value in being able to access a Windows session from your iPhone? I tend to think not, but it would be an interesting discussion. Speak up in the comments.
  • Duncan points out that the issue with adding NICs to a team and keeping them all active—the workaround for which required editing esx.conf—has now been fixed in ESX 3.5 Update 3. It’s now possible to add NICs using esxcfg-vswitch and there’s no need to edit esx.conf. Excellent!
  • If you haven’t yet checked out Leo’s Ramblings, go give it a look. He’s got some good content. It’s worth subscribing to the RSS feed (I did).
  • Rick provides a helpful tool to resolving common system management issues with VMware Infrastructure. Thanks, Rick!
  • Regular readers may recall that Chad Sakac of EMC and I had a round of VMware HA-related posts a few months ago (check out the VMwareHA tag for a full list of VMware HA-related posts). As part of that discussion there was lots of information provided about Service Console redundancy, failover NICs, secondary Service Console connections, additional isolation addresses…all sorts of good stuff. Duncan joined in the conversation as well with a number of great posts, and has been keeping it up since then. His latest contribution to the conversation is a comparison of using failover NICs vs. using a secondary Service Console to prevent VMware HA isolation response. According to the documentation, using a secondary Service Console can help reduce the wait time for VMware HA to step in should isolation actually occur. Good stuff, and definitely worth some additional exploration in the lab.
  • As a sort of follow-up to the discussion about using NFS for VMware, this VMware Communities thread has some great information on why the NFS timeouts should be increased in NetApp NFS environments. If you’re like me, you like to know the reasons behind the recommendations, and this thread was very helpful to me. Let me also add that we’ve recently started recommended to customers to increase their Service Console memory to 800MB when using NFS, so that might be something to consider as well.
  • Need to change the path of where Update Manager stores its patches? Gabe shows you how here.
  • Eric Gray of VCritial explores the question: what would things be like without VMFS? Well, as he states, you can just ask a Hyper-V user, since Hyper-V doesn’t (yet) have a shared cluster filesystem. Yes, that will change in 2010 with Shared Cluster Volumes in Windows Server 2008 R2 and Hyper-V 2.0. I know. Or you can just add Melio FS from Sanbolic today and get the same thing. This is not anything new to me; I discussed this quite extensively here and here. Now, what would really be interesting is for VMware to work with Sanbolic to co-develop a more advanced version of VMFS that eliminates the SCSI reservation problems…
  • Need a nice summary of the various network communications that occur between different components of a VI3 implementation? Look no further than right here. Jason’s site is another one that’s worth adding to your RSS reader.
  • If you really like living on the edge, here’s a collection of some RPMs for VMware ESX 3.5. Keep in mind that installing third-party RPMs like this is not recommended or supported…
  • Andy Leonard picked up the DPM video by VMware and is looking forward to DPM no longer being experimental. What he’d really like, though, is some feature to move his VMs via Storage VMotion and spin down idle disks. Andy, I wouldn’t hold my breath.
  • If you are a Fusion user (if you own a Mac and need to run Windows, you should be a Fusion user!), this hint might come in handy.
  • Eric Siebert has a good post on traffic flow between VMs in various configuration scenarios—different vSwitches, same vSwitches, different port groups, same port groups, etc. Have a look if you are at all unclear how VMware ESX handles traffic flow.

That does it for this round. Speak up in the comments if you have any interesting or useful links to share with other readers. I’d also be interested in readers’ thoughts on the whole Citrix on the iPhone discussion—will it really bring any usefulness?

Tags: , , , , , , , ,


  1. Andy Leonard’s avatar

    Thanks for the link – the comment about spinning down the idle drives was tongue in cheek – but I think you knew that already. ;) Mostly, I was pointing out (obliquely) that my storage systems, while also shared with many other systems, draw far more power than my VMware machines.

    But, hey, when DPM is ready, if it works for us, we’ll be using it. And I also wouldn’t be surprised if one of the big storage vendors someday had something that integrated with DPM… but I agree with your advice not to hold my breath.

  2. Dejan’s avatar

    Why do you recommend increasing Service Console Memory when using NFS?
    Could you explain the technical reason? Is there any kind om extra buffers or something like that?

  3. slowe’s avatar


    Great question! From what I’ve been able to gather, the Service Console helps to manage the VMkernel NFS connections, so increasing the memory for the Service Console provides more RAM for managing those NFS connections. In some cases, we’ve had VMware Support tell our customers that a lack of memory in the Service Console is the source of some of the NFS strangeness they might be experiencing. After having several customers told by VMware Support to bump the Service Console memory to 800MB, we decided to incorporate this into our standard configuration for NFS installations.

  4. slowe’s avatar


    Yeah, I knew you were joking, but it would be nice to have a storage vendor pay more attention to the power all those spinning drives suck down.

  5. Dejan Ilic’s avatar

    You have smaller disksolution vendors doing the “spin-down-unused-disks” already (notably NexSan with MAID). Check out

    But I wonder how that would work with consolidated systems that require lots of IOPS. You would probably have to use storage tiering (on block level) to be able to spin down disks at all.

Comments are now closed.