One of the biggest mistakes solutions providers make when implementing VMware Inc.'s vSphere is not reviewing the vSphere requirements for hardware and implementation. Having incompatible hardware can lead to a number of issues. Your customers may not be able to use some features properly, and they may not be eligible for vendor support. Find out why vSphere expert Eric Siebert thinks checking vSphere requirements for hardware is crucial to a successful implementation in an interview with SearchSystemsChannel.com's Pat Ouellette.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Pat Ouellette: Are there vSphere requirements and hardware compatibility issues that a solutions provider needs to be aware of before implementation?
Eric Siebert: Yes. That's probably vSphere's most critical and frustrating area. VMware publishes a hardware compatibility guide that basically tells a person that's buying hardware exactly what hardware they support with vSphere. And it's a pretty limited set of hardware.
There are two reasons for the hardware compatibility guide. The first reason is that vSphere requirements cover certain drivers for certain hardware. Therefore, if you get hardware that's not on that list, vSphere might not even see it and you won't be able to use it. The second reason is for support. VMware certifies all the hardware on that compatibility guide to ensure that it will work with vSphere requirements and that vSphere won't crash. Most operating system crashes are due to drivers, and [VMware] wants to have the vendor do a certification to ensure that [the hardware is] stable and will work with vSphere.
Some other challenges with vSphere requirements are that certain features in vSphere require specific hardware. If you don't get the right hardware, you may find that you can't use that feature. Fault tolerance (FT) is a good example of that. FT only supports very specific processor families. Therefore, between Intel and AMD, you have to look at the published guides from VMware to see if that server you buy has a CPU that will support FT. Other features, like VM DirectPath, require a specific processor and chipset function like the Intel VT-d and the AMD IOMMU. If you don't have that featured in your server, you won't be able to use the VM DirectPath feature.
Solutions providers must also remember that vSphere requirements include 64-bit hardware. It no longer supports 32-bit hardware. So your customers must have 64-bit hardware, and to properly use vSphere, they must also have the AMD virtualization and Intel virtualization technology extensions in their servers.
When looking for hardware, it's best to go through the compatibility guide because it doesn't just cover the server. It covers storage adapters, network adapters and storage. You have to go through [the guide] and check that every component you bought for the server is listed, because you may find you bought all the hardware and can't use it because it's not supported by vSphere. You'll have to either return it or go out and buy something else, which is a costly mistake.
Ouellette: What other factors does a solutions provider need to take into consideration when planning a vSphere deployment?
Siebert: Sizing is critical. Solutions providers need to make sure they size a host properly for the applications that they're going to put on [it]. There are tools that you can use when sizing a physical and virtual environment, such as VMware's Capacity Planner tool or its Guided Consolidation tool, that actually go out and measure the existing physical servers and their workloads. Using these tools, you'll actually get recommendations on the types of hardware and licenses you'll need for VMware to be able to run those workloads on the virtual host.
When looking at your customer's existing workloads, you need to do more than just look at them for 24 hours or a few days. You need to actually look at a longer period, like a month. You might find weekly or monthly cycles or processes that cause peak loads that don't happen all the time. You want an architecture environment to be able to handle both average workloads and peak workloads, because once you start combining applications onto single physical hardware, they're all going to be competing for resources. Solutions providers should ensure they have enough of all the resources available for the virtual machine so that they're not creating bottlenecks in their customer's environment.
About the expert
Eric Siebert is a 25-year IT veteran whose primary focus is VMware virtualization and Windows server administration. He is one of the 300 vExperts named by VMware Inc. for 2009. He is the author of the book VI3 Implementation and Administration and a frequent TechTarget contributor. In addition, he maintains vSphere-land.com, a VMware information site.