Windows Server 2008 R2 and Hyper-V Virtualization

Two years ago we were counting racks in the server room. Twenty-eight physical servers, most running at fifteen percent utilization. Today we have six, and more services are running on them than before. The key to this transformation was Windows Server 2008 R2 and its Hyper-V role. Here are our experiences from deployments at mid-sized Czech companies.

Why Hyper-V and Not VMware?¶

Let’s start with the most common question. VMware vSphere is undoubtedly a mature product with a longer history. But a VMware Enterprise Plus license costs significantly more than an entire Windows Server 2008 R2 Datacenter license — which includes Hyper-V at no extra cost. And for most of our clients — companies with hundreds, not thousands, of virtual machines — the feature set of Hyper-V R2 is more than adequate.

Moreover, if the client already runs Active Directory, System Center, and other Microsoft technologies, the integration is natural. SCVMM 2008 R2 (System Center Virtual Machine Manager) gives us centralized management, VM templates, a self-service portal for developers, and library integration. No additional vendor lock-in — these are all licenses the client already has.

Live Migration — Finally, Zero Downtime¶

This is probably the most important feature that R2 brought compared to the original Hyper-V in Windows Server 2008. Live Migration allows you to move a running virtual machine from one host to another without any downtime whatsoever. Zero downtime. For production servers, this is critical — you can perform maintenance on host servers while they’re still running.

The prerequisite is, of course, shared storage. In most deployments we use iSCSI SAN — specifically Dell EqualLogic solutions, which offer an excellent price-to-performance ratio. FC SAN is certainly an option, but for companies with dozens of VMs, iSCSI over 10GbE is more than sufficient.

One important note: Live Migration in Hyper-V R2 can only migrate one VM at a time. This is a limitation compared to VMware vMotion, but in practice we’ve rarely encountered it. Most of the time you’re moving servers in a planned, sequential manner.

Failover Clustering¶

High availability is a key requirement for our clients. Windows Server 2008 R2 offers fully integrated failover clustering that works excellently with Hyper-V. Configuration is straightforward — two to sixteen nodes in a cluster, shared storage, and you’re done.

Cluster Shared Volumes (CSV) are another R2 innovation that significantly simplifies management. Previously, each VM had to have its own LUN on the SAN. With CSV you can have multiple VMs on a single shared volume, access it simultaneously from multiple nodes, and Live Migration works without LUN ownership switching. In practice, this means less work for the storage admin and easier scaling.

Consolidation in Practice — Case Study¶

For a client in the manufacturing sector, we consolidated infrastructure from 34 physical servers down to 4 Hyper-V R2 host servers. The original servers were a mix of various generations of HP ProLiant and Dell PowerEdge, running Windows Server 2003 and 2008, file servers, print servers, SQL Server 2005, internal web applications, and domain controllers.

The migration took three months. We used SCVMM for P2V (physical to virtual) conversions of older servers and built new VMs from templates. The result: power consumption in the server room dropped by sixty percent, management is centralized, and backups via DPM 2010 run against VM snapshots — no agents inside, no complex configuration.

Networking and VLANs¶

Hyper-V R2 introduced an enhanced virtual switch that supports VLAN tagging directly at the port level. For enterprise deployments, this is a must — you separate management traffic, storage traffic, and client traffic into separate VLANs. We recommend a minimum of four network cards per host: one for management, one for Live Migration, one for CSV heartbeat, and one (or more) for VM traffic.

NIC Teaming in R2 unfortunately still depends on NIC vendor drivers — unlike VMware, which has its own solution. In practice, this means you use Intel ANS or Broadcom BASP depending on the cards in your server. It works reliably, but configuration requires a bit more attention.

Monitoring and Management¶

For monitoring, we use System Center Operations Manager 2007 R2. Management packs for Hyper-V are available for free and cover everything important — host health, VM CPU and memory utilization, replication status, and cluster resource health.

For capacity planning, Performance and Resource Optimization (PRO) in SCVMM has proven effective — based on data from Operations Manager, it recommends VM migrations between hosts. It’s not a full DRS like VMware’s, but it’s sufficient for our deployments.

What’s Coming — Windows Server 8¶

Microsoft has already hinted at what the next version of Windows Server will bring. Hyper-V 3.0 is expected to support up to 32 virtual processors per VM (currently a maximum of 4), Hyper-V Replica for disaster recovery without shared storage, and native NIC Teaming in the operating system. Those are exactly the things we’re missing today and why some clients still choose VMware.

We’re also watching developments around private cloud and System Center 2012. Microsoft is clearly heading toward Hyper-V being not just a hypervisor, but part of a comprehensive cloud platform. For our clients who aren’t yet considering the public cloud, a private cloud on their own infrastructure is the logical next step.

In Conclusion¶

Hyper-V in Windows Server 2008 R2 is not the most advanced virtualization platform on the market — that’s still VMware vSphere. But for most mid-sized Czech companies, it offers an excellent balance of functionality and cost, great integration with existing Microsoft infrastructure, and sufficient performance and reliability for production deployments. And that’s exactly what our clients need.