May 16, 2012

Part 5.3 - Clustering and Hyper-V Awesomeness

One of the challenges I can already see with all that is new and good when building out my private cloud is – which amazing clustering storage ‘thing’ should I use to enable my cloud:

  1. SMB Direct
  2. Cluster Shared Volumes
  3. Traditional Block Storage
  4. Shared Nothing

I don’t have an answer yet, but I do know that there are a variety of tools that SMB 2 support (discussed previously) bring to the table.  Additional stuff is outlined here.

Within a single cluster, I can now have 4,000 virtual machines.  Excessive?  Depends on how many nodes you have or how big those nodes are.  Regardless, that’s a lot of VMs!  The nice thing is that I can now perform large amounts of tasks across my VMs collectively (bulk migrations, updates, etc) and Clustering Services is smart enough to keep things up and running by queuing things up when the load gets high, powering off lower priority VMs (called preemption), etc.

NUMA aware virtual machine placement.  Non uniform memory access is important to understand and something I am still working on.  Essentially, it is a way for Hyper V VMs to be aware when NUMA is in use on a Host and allows the VM (and placement engines) to evaluate more fully whether a VM will run successfully on a host.  Cool underpinning.

Automatically draining a cluster node – if you consider previous article on cluster aware patching, this makes sense – when you put a host node into maintenance mode to patch it, be smart – move the VMs off to other nodes and then patch.  Then, perform the next host node, etc.  This also extends to the VMs themselves which is a bit freaky.

These are all underpinning things, but speak to the power of what the clustering services will be able to do for me.  Next time, I’m going to return to my storage question and talk more about Cluster Shared Volumes and what Windows Server 8 did to them.