I work as a Technical Specialist for an Independent school in the UK, and I have been kindly asked to provide a post to share my experience of using Windows Server Hyper-V. Coincidentally it’s almost 4 years to the day that I visited Microsoft to talk about virtualizing our systems, so this has given me a chance to reflect what has happened over that time…
In the summer of 2008 I was given the task of virtualizing our entire server environment. Our physical servers were loaded up with all sorts of services, working on them was a nightmare, and the requests for more was growing. Obviously I was aware of the main player in this field, but I wanted to keep an open mind. So, I made a call to our Microsoft support company, and they got me in touch with Matt McSpirit, who at the time was a Microsoft virtualization evangelist.
To be totally honest, I didn’t expect very much. I mean, why would anyone in Microsoft spend any time talking with me about virtualizing our servers? For someone in my position it’s easy to think that direct contact with Microsoft is really not possible. To my surprise, Matt was extremely approachable, so much so in fact that he invited me down to Microsoft TVP Reading for a chat.
Autumn 2008: The meeting with Matt went very well, and I can tell you that he certainly knows his stuff. It wasn’t rushed, it was well balanced, and considered all angles, not just Microsoft’s. This was definitely not a sales meeting, more a meeting of minds. But the excitement over Hyper-V was obvious. In fact it was infectious, and this was the start of our virtual journey with Hyper-V.
Our Virtual World
Hyper-V, at least for us, just worked. We started with a 3 node failover cluster, sitting on top of an EVA SAN. I soon added in System Center Virtual Machine Manager (SCVMM). With that, creating and managing VMs couldn’t have been much simpler.
But life wasn’t without its gotchas. One morning I arrived to work to find the whole system down—that was all our VMs, cluster, everything. During the night there had been a prolonged power-cut, and the system had shut itself down after our UPS had drained. The worrying part was that the cluster refused to start. This was due to the fact that the DCs now existed as VMs inside the cluster, no DCs = no trust, no trust = no cluster, and no cluster = no DCs. There was a 10 minute period that felt like an eternity, where I was just staring at the hosts thinking, “What the hell do I do now?”! Fortunately, I had only just P2V’d the DCs the week previously, so after quickly firing up of the old machines we were back up and running. Suffice to say since then we’ve always kept one physical DC.
However, not one issue I have come across has caused any of our systems to be down for more than half an hour during working hours. Over the last 4 years Hyper-V has provided a reliable, stable environment. In fact I don’t even worry about it.
The cluster now has 4 nodes, and with increased memory it contains 50+ servers. There are also have other hosts for backup, and other solutions not suitable for the cluster. To give you an idea, here are a few services we currently provide in our Hyper-V environment:
- Microsoft SharePoint 2010
- Microsoft Lync Enterprise
- System Center
- Virtual Machine Manager 2008 R2
- Operations Manager 2007 R2
- Service Manager 2012
- Configuration Manager 2012
- Data Protection Manager 2012
- Microsoft Exchange 2010
- SQL Server
- Threat Management Gateway (TMG)
- Remote Desktop Services (RDS)
Growth and the Future
As a business we began our virtual journey roughly at the same time Hyper-V began its, and we have grown together. As more of our critical servers became virtualized, along came live migration in Windows Server 2008 R2 Hyper-V to reduce interruptions when moving machines. As the VM count grew, so came along dynamic memory in SP1 to help utilize resources more efficiently.
The only minor limitations I have come across would be:
- Windows Server 2008 R2 allows 4 Virtual Processors
per VM. The Hypervisor can safely support 8 vCPUs per core, and we have 24 cores
per host. That’s a total of 192 vCPUs! With the current limitation I’ll run out
of memory far sooner than I’ll run out of vCPUs
- 1 Live Migration at any one time
- VLAN Management
- Dynamic balancing of the cluster
- SAN Management
Now as we move towards Windows Server 2012, we start to see these limitations being stripped away. One of my next projects is to completely rebuild our current cluster, to split it between two locations, with brand new host servers, and add in a clustered iSCSI SAN. Crucial to the smooth running of this will be the new features in Windows Server 2012 and System Center 2012 Virtual Machine Manager.
In fact, those last 2 points are more about my time than anything else, and two features of SCVMM 2012 I’m looking forward to checking out is; dynamic optimization of the private cloud, and SAN integration. Private Cloud means different things to different people, but to me essentially Private Cloud = My Time Saved.
VLANs: Our network relies on them. Up until now, it’s all been done at the port level, which really locks down the NICs on the host. I’d love to forget about VLANs on the port—in fact I’d like to get rid of teams on the host and simply let the VMs use as many NICs as are available and let them set the VLAN they require. Sounds as though this might also be possible.
Live Migration: Because I work for an Independent school with students from all around the world, we have to run a 24/7 zero downtime system. Regular maintenance on the hosts is tricky enough, without the pain of live migrating each VM one by one. So I was pleased to hear restrictions on Live Migration have gone with Windows Server 2012. But I don’t want to spend my time patching the host., I want to do more interesting things, so being able to “Patch my Fabric” with System Center 2012 Virtual Machine Manager is again another great addition.
As for vCPUs and memory, well it’ll be some time before we reach the new limits.
One thing I have not had time to really investigate just yet is the new VHDX format. But from what I’ve read so far on performance and capacity it sounds extremely promising.
Looking back, I guess we were early adopters of Hyper-V, and sometimes being on the bleeding edge has its downfalls. Being a school we are privileged to have Microsoft’s amazing educational pricing, and you could say that Hyper-V from that standpoint was a no brainer. But also, we were starting from nothing. Had we gone with its main competitor, would we have utilized all the features it provided over Hyper-V? Would we have got value for money? Possibly, possibly not. The knowledge I gained from Matt McSpirit at the very start, meant I knew Microsoft had a strong product and was determined to develop it further, and so far that has held true. I believe I made the right choice, and 2013 is going to be an exciting/busy year.