Virtualization may seem like old hat now that attention is shifting toward the cloud, but the fact is that the virtual layer is and will remain an integral component of enterprise architecture no matter which direction it takes.
Of course, the same could be said of plain-old physical infrastructure, which, as it turns out, has a significant impact on the cost containment and data flexibility goals that virtualization and the cloud are intended to achieve.
For instance, virtualization is initially deployed as a means to consolidate server infrastructure. But according to The Register's Alan Stevens, that can only be accomplished if the great majority of your processing power sits idle most of the time. And believe it or not, utilization rates may not be as low as you suspect considering how easy it is to sneak database or Web server apps onto a server undetected. In the case of older hardware, virtualization might also be of limited value depending on whether processors can support the latest hypervisor releases. Either way, the need for clear visibility into server, network and storage infrastructure is a must before any virtual deployments are undertaken.
In many cases, some hardware upgrades will be required before the first virtual machine is powered up. As InfoWorld's Paul Venezia noted recently, speed is no longer the top priority when choosing CPUs. Depending on your needs, a greater number of 1.7 GHz Opterons would outperform fewer 2.93 GHz Westmeres. Remember, few subsystems can keep up with the fastest processors, so they have to wait while the data trickles in anyway. As more software becomes multi-threaded to accommodate multicore environments, a larger number of slower cores will actually get the job done faster.
Fortunately, it seems that many enterprises are getting a better handle on what virtualization can and cannot do. According to Symantec, there is only a 4 percent gap across the board between enterprises' expected and realized goals for virtualization. The biggest disappointments tend to come in scalability and reduced capital and operating expenses, while 85 percent of enterprises report satisfactory results when it comes to actual performance.
This all points up to the fact that nothing in the enterprise - whether real or abstract, local or remote - exists in a vacuum. Optimal performance can only be achieved if all aspects of the environment are in top working order and are in tune with each other. And that requires not only a well-defined data infrastructure, but the ability to peer into its constituent parts to see exactly what is going on.