Have You Any Constructive Feedback for Microsoft’s Virtualisation Teams?

I’m off to spend a few days with the virtualisation product teams in Redmond (Microsoft HQ) at the annual MVP Summit in a few weeks time.  This is our opportunity to learn some new stuff and to give feedback to the program managers of the various products.  In my experience with MS, the virtualisation folks have probably been the most responsive teams to feedback.  And that leads me to the point of this blog post.

What constructive feedback have you got for them?  Are you having issues with CSV?  If so, what are they?  Is networking proving to be a challenge?  Are you having difficulties with support issues?  What’s your experience been with Linux integration components?  Is there anything else on your mind (keep it constructive!)?

Post a comment on this blog post if you’ve got anything that you’d like me to relay to Redmond.  There are no promises but I’ll pass it on.  I did the same last year, and the folks there were interested – in fact, in some cases they were thinking the same things as you. 

And this year … well … it’s no secret that Windows “8” will RTM around mid 2012.  That means a new version of Hyper-V is in the works for the server.  And thanks to the MS France blunder, we know that a version of Hyper-V for the desktop OS is also on the way (after years of begging for it).  This is the year to make an impact on design decisions.  2012 will be way too late – we know that substantial changes don’t happen once a product reaches beta stages.

Please follow and like us:

17 Replies to “Have You Any Constructive Feedback for Microsoft’s Virtualisation Teams?”

  1. NIC teaming support from Microsoft would be welcome! Even the possibility to team different (vendor) nic’s. Last but not least a distributed virtual switch…

    My2Cents

  2. OK, I’ll kick things off:
    – Find a way to do storage level backups of CSV without any redirected I/O. Probably requires a granular VSS snapshot.
    – Granular VSS snapshot. People want the ability to backup a subset of VMs on a CSV without snapshotting the entire LUN
    – More granular subdivision of organisation or delegated administration in VMM. In other words, host groups only give us a physical representation. We want a logical representation so that different admins don’t need different compute clusters.
    – It’s a commonly requested one: NIC teaming that is done by the virtual switch, not by 3rd parties. Even if just for fault tolerance, e.g. the vSwitch detects path failure and redirects to another NIC.
    – Some middle ground on support for Linux distros. Be nice to see supported ICs for CentOS and Ubuntu. Some customers are going straight to VMware because of lack of “support”.
    – Simpler and automated Linux IC installation. We need a way to automate this. Larger installations, e.g. public/private cloud, will not tolerate the existing system. Most public hosters have 50%+ Linux (usually CentOS in these parts).

  3. – global vswitch
    – *BSD support
    – multi-server config editing using HV manager (GUI, i.e. without powershell)
    – HV manager generating post-config powershell cmdlets for copy/paste

  4. I like the things listed above, and thought I would add a couple:
    -Improved ability to manage guest storage in VMM.
    -Ability to change the the default so that Time synchronization ISN’T offered.

  5. Oooh, Michale’s post reminds me …. CSV visibility in VMM and in OpsMgr monitoring. For example:

    – I want to be able to see health, free space, etc, in monitoring and reports.
    – I want to be able to control placement of VHDs on CSVs, assuming I’ve got tiered storage (private cloud scenario).

  6. What most of these really are (the ones related to CSV anyway), are requests for a real cluster filesystem, not a CSV bolt-on on top of NTFS. Hyper-V really needs a full Cluster file system then all the CSV requests would work. This would also allow you to live migrate more than one VM at once I believe. This is also my biggest request, my others would be

    – Full IC support for Linux. This measn all the features of the IC’s for Windows, not just bits and pieces.
    – BSD Support, thats also got everything, not just storage and NIC
    – Ability to shrink VHDs. Ever done a P2V from a machine with a 500GB drive with 40GB used, it’s enough to make you cry.

  7. Hyper-V needs:

    1) A real clustered file system. MS should acquire Sanbolic and incorporate it into Server 2013.

    2) MS needs to use the new array SCSI commands VMware pioneered (which can be used by anyone) to offload block copies, metadata locking, and block zeroizing.

    3) A Cisco virtual distributed switch (aka Nexus 1000v for Hyper-V).

    4) Need to add array friendly thin provisioning storage reclimation. Basically the SATA “trim” command for enterprise arrays, so thin volumes stay thin when you delete data within the VM.

    5) Massively reduce server core’s footprint so that it’s vastly thinner and leaner. Make it more along the lines of ESXi. Tiny purpose built hypervisor, not something bolted on a general purpose OS.

    6) Ability to do 4 -8 concurrent live migrations per host.

    7) Have a separate WHQL for Hyper-V certified drivers that significantly raises the bar for quality. Don’t block non-certified drivers, but offer a very rigorous testing certification for hypervisor drivers. Can’t tell you how much I hate the HP NIC drivers.

  8. I think for me the biggest improvements could be NIC teaming that is done by the virtual switch, not by 3rd parties and a global vswitch. Also the way network components are displayed could be made more visual….

    Another thing, maybe more depending on changes in Windows Server Core… an easier way to manage NIC’s (for Hyper-V) in Server Core. Especially when using teaming software, it’s really hard to currently manage and document this. Let alone pass this knowledge on to customers in an understandable fashion.

  9. So what I am hearing is that I should cancel the 3 hour “Cooking with Ben” session I had planned? That is a shame – cause I am getting really good at making omelettes…

    I guess we can talk about some of this stuff instead 😉

    Cheers,
    Ben

    (the guy at Microsoft who will be on the receiving end of the “constructive” feedback).

  10. Speaking of Live Migration: The ability to use redundant Live Mig Networks simultaniously so when multiple Live Migs are an option can be spread over those NICS. Could be handy when you daint’ have 10Gbps.

    Others: get the Hyper-V Manager an VMM synced in terminology, action/results. What I’ve seen of VMM 2012 they are integrating anything cluster related into it. How far will they go with this in Hyper-V Manager?

    I’d also like it if they supported CSV in Windows Backup.

  11. After LM’ing 8 VMs using the maintenance option in VMM I’d love the it to give you the option of LM’ing them back again.

    Better Time functionality needs sorted within IC’s for VM DCs. Hosts getting their time from child VM’s is not good, something should be put in place so manual intervention is not required.

  12. While I can understand the reasons to re-use NTFS, Microsoft’s answer to the cluster file system needs to be revisited. Redirected mode for CSV certainly needs more thought, both in terms of practicality for backup (HW/VSS) and the mechanism through which it maintains state. Don’t even think about touching the preferred network for CSV communication (even if you have multiple), the time it takes to ‘fail’ over to the next takes too long, potentially having disastrous effects.

    Implementing ‘traditional’ networking technologies such as adapter teaming is still a pain, Microsoft still stand behind the ‘supported by your hardware manufacturer’ statement; however the reality is that the industry demands this functionality as standard, with cohesive support. Better support for these basic features in the operating system must be realised before Hyper-V can be taken seriously as both an enterprise production hypervisor or private/public cloud solution. More advanced functionality such as VMQ (Virtual Machine Queues) is not new, and can be referenced in the HSN Deployment Guide, but enabling it for Hyper-V in my experience has always ended in tears.

    SCVMM? Sometimes I wonder if it’s connected to a different cluster. SCVMM seems to have its own (incorrect) opinion as to the state/health of the cluster. To be honest, I’d prefer to just use the failover cluster snap-in and powershell any day.

  13. Hi,
    I would vote for some optimization in storage area. Mainly:
    – I’ve been using diff disk for testing environment, but you are unable to control this from within VMM (for instance: one parent, 100 child, now create snapshot on all of them – you will receive “object is in locked state”).
    – anything from http://www.vmutil.com
    – some kind of storage voodoo that would improve speed:)
    (+ cluster support for NIC teaming and TOE)

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.