Due to time constraints, I gave up on the Server Core installation and ran Hyper-V on a full Server 2008 installation using the local console for management. After installing Server 2008, I installed the Hyper-V role using Server 2008’s Server Manager. Not counting the failed Server Core installation attempt, it took three reboots before the server was fully set up.
Figure 2 shows Hyper-V’s management console. Theoretically, multiple Hyper-V server instances can be managed in the left pane of Hyper-V Manager, although I couldn’t connect to any remote servers. Selecting a server instance displays that server’s VMs in the Virtual Machines pane. You can then manage the VMs by right-clicking them and selecting options from the context menu. I found Hyper-V Manager to be less functional and less well thought out than ESX Server 3.5’s management console. I thought that the Snapshot pane in particular was poorly positioned, and it seemed that everything was forced to fit inside the generic Microsoft Management Console (MMC) 3.0 framework. Although Hyper-V Manager doesn’t show detailed server performance information, basic VM management using the management console was functional and easy.
SQL Server VM Performance
Performance is where the rubber meets the road. To compare the performance of ESX Server 3.5 and Hyper-V, I created four VMs, each of which was running Server 2008 Enterprise Edition configured with 512MB of RAM running SQL Server 2005 Enterprise Edition SP2. For each platform, I used the default virtual hard disk configuration options. Four Server 2008 instances can be simultaneously active as VM guests with no additional Server 2008 licensing costs. (I’ll discuss licensing in more detail later.)
All of my testing was conducted on an HP ProLiant ML370 G5, which is a rack-mounted 4U server. My test unit included two Quad-Core Intel Xeon processors running at 1.86 GHz on a 1066 MHz front-side bus. The ProLiant ML370 G5 came equipped with 8GB of RAM and eight 72GB 15,000 RPM drives configured as a RAID array.
For the test workload I used 27 different queries running against the sample AdventureWorks database. Although the bulk of the workload was data retrieval, the batch also contained a couple of computational loops and four SELECT INTO statements to add some computational and data modification operations. A think time of three seconds was inserted between each database interaction.
The workload was executed from four physical client machines, each of which used Microsoft’s sqlcmd utility to launch the workload against all four SQL Server systems running as VMs. Then the client paused for three seconds. This process was repeated 10 times, creating a total of 40 running jobs from each of the four clients where each client was running 10 separate connections per server. This test revealed how each of the virtualization platforms was able to respond to increasing workloads and to the total workload with all four virtualized SQL Servers actively servicing 40 queries. The final performance measurement that I used was the average time required to complete the measured workloads. The tests were run multiple times and the results were averaged. Web Figure 2 compares the overall performance of the two platforms.
As you can see in Web Figure 2, the results were surprisingly close, with ESX Server 3.5 providing better performance under the tested workload. Although ESX Server 3.5 edged out Hyper-V by 4 percent in the performance tests I ran, Hyper-V made a competitive showing in spite of the fact that it was prerelease code. I expect the final release will provide very similar performance results. I also expect that the glaring remote management problems will be corrected in the final release. However, I don’t expect Hyper-V’s management console to be improved to match ESX Server 3.5’s Virtual Infrastructure Client. You should bear in mind that although these results were representative of the test workload that I ran, it’s possible that different workloads and environments could produce varied results.
Virtualization and Licensing
Price is another area in which there’s a significant difference between ESX Server 3.5 and Hyper-V. ESX Server 3.5 has always been a chargeable product and is the staple of VMware’s product line. In my comparison of each of the product’s licensing and pricing structures, I used VMware Infrastructure Foundation 3, which includes the following components:
- ESX Server 3.5
- VMware Virtual Machine File system (VMFS)
- VMware SMP
- VirtualCenter Agent
- VMware Consolidated Backup
- VMware Update Manager
In contrast, Hyper-V is included in Server 2008, making it essentially free to organizations running Server 2008. Because Hyper-V is a 64-bit technology, you must have x64 hardware and use one of the x64 editions of Server 2008. Hyper-V is included in Server 2008 Standard Edition x64, Server 2008 Enterprise Edition x64, and Server 2008 Datacenter Edition x64.
Microsoft does make three versions of Server 2008 that don’t include Hyper-V (aptly named Server 2008 without Hyper-V), and the price difference is negligible at only $28. Hyper-V also isn’t included in Windows Web Server 2008, Windows HPC Server 2008, or Server 2008 for Itanium-based Systems.
The pricing for the two tested configurations is shown in Web Table 2. Note that this table uses retail pricing, whereas most businesses use Microsoft’s volume licensing.
One important point to be aware of is that Server 2008 Enterprise Edition and Windows 2003 R2 allow for as many as four active virtual Windows instances with no additional costs. This licensing is the same whether you use Microsoft or VMware virtualization products. In addition, Server 2008 Datacenter Edition and Windows 2003 R2 allow an unlimited number of virtual Windows instances no matter which virtualization platform you’re using. You can also run an unlimited number of SQL Server 2005 Enterprise Edition instances on VMs no matter which virtualization technology you’re using. The end result is that the inclusion of Hyper-V with Server 2008 makes Hyper-V a less expensive virtualization option than ESX Server 3.5.
The Bottom Line
My testing revealed that both Hyper-V and ESX Server 3.5 deliver excellent levels of performance. For midsized businesses, especially those organizations just getting into virtualization, I found Hyper-V to be compelling. Being bundled with Server 2008 makes the price point very attractive and the Windows-based management tools make it easy to use. However, ESX Server 3.5’s better performance, more mature management tools, and Virtual Infrastructure 3 management suite provide a more feature-rich platform for large organizations.