Creating a cloud service in Azure

with 2 Comments

Sometimes, it’s the little things that matter the most.

Creating a cloud service is more than just creating a cloud service. Have you ever wondered why you can’t scale beyond A-series? That is because the first VM you deployed to the cloud service is in the A-series. If you do this, Azure will create your cloud service on an old version of the hypervisor fabric. The old version doesn’t support D-series or higher.

a-series

But if you instead start by deploying your first VM as a D-series, you will be able to scale a bit more. You can of course scale down to A-series immediately, but at least now you have the option to scale if you want. You can now deploy VM’s as A-series in the same cloud service and later scale to D-series if needed.

d-series

In my case G-series is not available in the north Europe datacenter, but it would also be listed here if I deployed it to a datacenter that supported G-series. If you are planning to use a big G-series machine you should deploy this VM first. This is because your large G-series VM will be deployed on a new hypervisor, and your upcoming VM’s will be placed (physically) as close as possible to this VM.

But there is also one more thing that is kind of important. If you deploy to an old hypervisor-cluster, when Microsoft is upgrading the fabric. Your VM will be rebooted by Microsoft. If you instead use the new hypervisor-cluster, your VM will be paused and then resumed during fabric maintenance.

Follow Viktor Gilbertsson:

Private cloud, public cloud, hybrid cloud, community cloud. Kind of a cloud enthusiast. IT professional.

2 Responses

  1. Johan Classon
    | Reply

    I had no idea. I what you write is correct, I think it is kind of sad that Microsoft does not keep their hypervisor-clusters updated! Great advice about pause/resume instead of reboot.

    • Viktor Gilbertsson
      | Reply

      Information is coming from an Azure TSP at Microsoft. Maybe I did not explain enough. Of course Microsoft is updating their hypervisors.That is why this phenomenon is happening. The older versions don’t support pause/resume. But they are still great hypervisors and also need maintenance, and that is why the reboot is needed.

      Microsoft is just like every other company in the world, they need to make money. Lets imagine the expected life time of a host is 3 years. During these 3 years new technologies will be invented. And Microsoft would probably go along with this, buying the next generation hosts that supports new technologies. But it would be to expensive to decapitate the “old” hosts.

      D-series is using SSD to create a temporary cache disk on your VM, while A-series use HHD.

      If you got the knowledge you can all ways move your VM’s to another cloud service! :)

Leave a Reply