The virtual skinny

The trend for virtualization isn’t all that new at all, as Tony Iams tells BM.

The concept of virtualization has been around for 30 years or more. Historically, before VMware entered the market, increasing amounts of operating system technology was related to managing resources in dynamic ways: adding hardware online (particularly in the UNIX space), advancing resource management, managing memory and CPU. When VMware entered the market, virtualization was viewed as the next phase of evolution for operating systems.

What is new is that virtualization is now available for industry-standard x86 hardware.
“That is what VMware accomplished, because x86 is inherently not suited for virtualization,” says Tony Iams, VP and Senior Analyst at Ideas International. “VMware figured out how to make that work and launched this market. Now competitors are coming in, and you have a dynamic marketplace.”

As more vendors enter the marketplace, it validates the future of the industry. VMware has company in the form of Microsoft and new independent players like Citrix (which acquired XenSource), Virtual Iron and SWsoft. This new activity has rejuvenated some of the virtualization efforts on other platforms as well. “That’s a solution for doing extreme consolidation of workloads. Microsoft is getting ready to launch its platform there. All this underscores the fact that it is going to be assumed that virtualization is available,” as Iams puts it. In the future, it is going to become a standard part of IT infrastructure.

What this means for data centers is that it changes the assumptions about how workloads should be managed. It will be taken for granted that you have the flexibility to run workloads where it makes the most sense to run them; based on performance requirements, based on temporal distinctions, different times of day, seasonally. Dependent on these factors, you may be able to move workloads around. Workloads will also have complete transparency in terms of hardware and geographical location. This gives administrators flexibility for planning data centers; they won’t have to overprovision anymore, which means use of those resources can be dynamically varied over time.

Optimizing virtualization
Utilization is a topic that is often discussed in this context. Historically, users would overprovision servers to ensure they had the headroom over time should it be needed. The reality was a lot of the time these servers with resources weren’t used, so sat there idly. “On x86, utilization is a major concern because you have server sprawl and large numbers of servers running at only a fraction of capacity,” says Iams. “Consolidation of those servers is a high priority. You can achieve impressive results by taking underutilized servers – which might be running at 10 percent capacity or less – and stacking them onto virtual machine hosts at ratios of five, 10, 20 or even more.” With virtualization there are fewer workloads that are critical, but they have more of a need for dynamically varying their resources over time.

Bearing this in mind, it is entirely possible to lower your cost in terms of hardware acquisition. “You’re buying fewer servers, which means spending less money on hardware. The hardware that you do buy is going to get much better utilization. You’re going to be using more of the hardware that you actually acquire,” says Iams. There is also an indirect cost reduction, as less energy is needed to run fewer servers: power and cooling needs are decreased.

While these costs go down, you might be surprised at the impact on the cost of management, or ‘soft’ cost. It is possible this might increase, as more money needs to be spent on training. There might be an incremental cost when introducing the virtualization technology, but at the same time, the operating system footprint that has to be managed isn’t reduced at all.

“Those operating systems have to be patched and provisioned and so on - cared for and fed, if you like,” explains Iams. “It’s just that they’re now running on a virtual machine instead of on a physical server.” So as virtualization is introduced, management tools should also be adapted to take this into account.

Users that do this tend to be more satisfied, because they get a quicker turnaround on their requests which in return allows organizations to respond to new business opportunities much more quickly. “Almost everyone agrees that this is a good way to do things. You might not save money, but it could help you make extra money because you’re responding much more quickly. It’s harder to put a price tag on what that’s really worth, but almost everyone agrees that it’s a positive step.”

A virtualized future
Virtualization is becoming a standard part of the infrastructure at almost every level. The leading x86 manufactures – Intel and AMD – are building virtualization hardware into their processors and are currently addressing CPU and memory virtualization. Not all applications are suitable for virtualization right now on x86. On other platforms the bottlenecks have been overcome, so UNIX systems and mainframes have no performance trade-off.

At the lower levels of the stack, you’ll be able to take virtualization for granted, Iams states: “This will mean shifting attention to the virtual infrastructure that you build on top of these virtualization platforms in the hardware and in the OS, which affects things like clustering and load balancing. All of that is required to be able to move workloads around at will. Again, you’ve started off with the premise that you’ll be able to move workloads, connect those workloads with resources that make the most sense based on their specific requirements. In order to do that reliably, you need to have the tools for controlling that virtual infrastructure.

“Almost every vendor I talked to is pursuing virtual infrastructure. Operating system vendors are working on this, as are the hardware vendors and traditional system management vendors: CA, BMC and so on. There are a profusion of start-ups pursuing this: I can think of probably 30 different solutions out there for dealing with virtual infrastructure, so there’s going to be a lot of competition in the short-term, then eventually consolidation. We’re still in a dynamic phase of that process, but that’s where a lot of the development is happening right now,” Iams says.

While a lot of the interest has been in server virtualization, it can also be applied to other parts of the infrastructure as well. One area of growth is in desktop virtualization. Virtual Desktop Infrastructure allows users to run desktop systems, host desktop systems on servers and then allow them to tap into some of the functionality on a remote basis through application streaming. There are a number of different ways to implement this, but all of them stay off of the client and onto the server in some form. This has cost benefits, because those desktop environments can be managed with better economies of scale once they’re hosted on the server.

Network virtualization is something that Cisco has been talking about lately, as Iams says: creating dynamic bindings between servers and the network infrastructure. He then points to storage virtualization, which is starting to gain traction. Today virtualization is being looked at in every part of infrastructure. There’s no doubt that in five to ten years time, it will be taken for granted: no better time than to acquaint your enterprise with this technology than now.

About Tony Iams
Tony Iams, VP and Senior Analyst, manages Ideas International’s System Software research program.

Tony focuses on evaluating and contrasting the features and functions of the leading operating system and virtualization technologies in use today. By applying a detailed analysis methodology, Tony identifies the strengths and weaknesses of current products and guides IT decision makers on the most effective system software technology choices. Tony’s research on virtualization has focused on the issues related to decoupling workloads from details about the underlying operating system and hardware platforms, so that resources can be matched with workloads more flexibly. The key virtualization technologies covered include logical partitions, resource management software, virtual machines, blade servers, provisioning software, and Grid computing.

Green computing
With or without virtualization, consolidating data centers and green computing are mandates in their own right. Virtualization is a powerful way to do consolidation, but it is not the only way, and it is not its sole use; virtualization has been applied successfully in branch offices, for example. There are other forms of consolidation: application or database consolidation. Virtualization is highly aligned with green computing because again you’re fundamentally reducing the hardware footprint with virtualization, and using less power as a result. At the same time, because you’re increasing the utilization of those hardware resources – depending on how the server is designed – it is possible that total power use is increased, because a server that’s running at 10 percent utilization consumes a lot less electricity than a server running at 90 percent utilization.

Then there are other aspects. Blade servers, for example, are often coupled with virtualization. Many users who are bringing in one technology are also looking at bringing in the other and often running them at the same time. Again with blade servers the use of power is being concentrated into a smaller space, which has an impact on cooling requirements.

Virtualization can potentially help with green computing initiatives, but there are many other questions that have to be answered with green than just the virtualization aspect. They’re aligned, but they’re not identical.