Executive Summary:
If you're running virtual machines (VMs) in a production environment and are having performance problems, InovaWave's DXtreme for Windows Enterprise could be just the product you’re looking for. DXtreme for Windows Enterprise has the potential to enhance VM performance by reducing the Average Disk Queue Length and providing a performance boost in transactions per second (tps).

Virtualization provides several great business benefits such as the ability to consolidate servers, increase business continuity, and streamline new server deployment. However, virtualization’s weakness has always been performance. Even with today’s virtualization-aware processors, virtual machines (VMs) simply don’t offer the same level of performance as a physical system. Often times, the level of performance limits where an organization can deploy VMs. For instance, if you have a missioncritical application that must meet a certain service level agreement (SLA) or have a heavily utilized enterprise-level translational database, you might find that these applications are unable to deliver adequate levels of performance when running in a VM.

InovaWave’s DXtreme for Windows Enterprise is a unique product that’s designed to enhance VM performance. DXtreme Enterprise works with both Microsoft Virtual Server 2005 and VMware’s Virtual Server products. It requires a minimum of 3GB of RAM, although 4GB of RAM or more is recommended. The product also requires 5GB of free disk space and runs on the x86 versions of Windows Server 2003 Standard Edition SP1, Windows 2003 Enterprise Edition SP1, and Windows 2003 Enterprise Edition R2 SP1. It’s important to note that the current version of the product doesn’t support the 64-bit versions of Windows.

Installing DXtreme Enterprise was quick and easy; the total installation process took only about a minute. This product is installed on the virtual server host system. At the conclusion of the setup process, you simply point DXtreme Enterprise to the specific VMs that you want to give a performance boost. Note that you can target all VMs or individual VMs. Because the product relies on a file-system driver, you must reboot the virtual server host system once the installation process is complete. DXtreme Enterprise is managed using the DXtreme Enterprise management console (Web Figure 1), which you can start by clicking Start, All Programs, InovaWave, VMX Suite DXtreme, DXtreme UI.

After the product was installed, I reconfigured the test VMs according to InovaWave’s recommendations, which included reducing the memory allocated to the VM. That’s right—reducing the VM memory. The product documentation, which was included as a PDF, recommends playing with various settings to find an optimum amount of memory. I chose the middle recommendation and reconfigured the test VMs down from 512MB of RAM to 256MB of RAM. Although you’re not required to reduce the memory, it is recommended because DXtreme Enterprise works by predicting disk access, and the lower memory values require more access because they force the guest OS to perform more paging.

I tested DXtreme Enterprise on an HP ProLiant ML350 with dual 3.2GHz Intel Xeon processors. This system included 6GB of RAM and a 320GB 15,000 rpm raid storage subsystem. The system was running Windows 2003 Enterprise Edition SP1 with Virtual Server 2005 R2. To ensure a consistent workload, the tests were performed using a single active VM that was running SQL Server 2005. I tested the SQL Server performance using Idera SQLscaler’s built-in Relation Database workload running against an AdventureWorks sample database. To get an idea of the performance difference, I first ran a set of three tests against the unmodified SQL Server VM using SQLscaler. Then I installed DXtreme Enterprise and reran the three tests. A subset of the representative test results is shown in Web Table 1.

DXtreme Enterprise did make a definite difference in the VM’s disk performance. The most significant difference was in the Average Disk Queue Length, which measures the number of disk requests that are waiting to be serviced. DXtreme Enterprise essentially found the information off disk more efficiently, which drastically reduced the Average Disk Queue Length. Aside from raw disk performance, database application performance typically boils down to transactions per second (tps), and my tests showed DXtreme Enterprise provided approximately a 20 percent performance boost in tps. Obviously, these results could change depending on the workload, but they do indicate that DXtreme Enterprise has the potential to enhance VM performance.

If you’re running VM’s in a production environment and are experiencing performance problems, then DXtreme Enterprise could be just the product you’re looking for. Not only can DXtreme Enterprise turbocharge VM performance, but it can also reduce how much memory the VM uses, enhancing the VM host.