Dario Lo Presti - Fotolia
In today's world of hardware abstraction, does the vendor of your server for virtualization matter anymore? For years, the answer has always been, "of course." We ran applications on the hardware we trusted, and many businesses were fiercely dedicated to those vendors.
However, virtualization's key feature in abstracting the hardware from the server operating system and application started marching us down a path where the server vendor often doesn't matter. However, many are still skeptical.
All the same under the hood
No matter how far we take abstraction, we must still admit that hardware matters. Most platforms will use chips from Intel or AMD and, in fact, most servers will have many of the same internal pieces. This does not mean quality doesn't matter, but because most of the major hardware providers have standardized on common components, most hardware platforms are remarkably similar. For this reason, server hardware choice becomes more of a selection of value added services, configurations and software rather than the hardware itself.
Features you don't need or won't use
Of course, this is not an idea that all hardware vendors are willing to embrace. Often the argument in favor of brand names is better thermal protection and information. Servers today, for the most part, comply with the Common Information Model, providing an open framework for gathering server operating details such as key temperature or fan speed values.
Does a server with more sensors lead to the collection of more valuable data? In an ideal world, the answer is yes. A management system and dedicated personnel should be able monitor this data 24/7 and take proactive steps to avoid problems and optimize performance. But we all know that the reality is personnel shortages and tight project deadlines make this level of monitoring a fantasy. Reality aside, vendors are more than willing to push these features as necessary for an ideal data center that is simply out of reach for many.
The other argument in favor of a specific server for virtualization is support for features such as dynamic cooling, which helps monitor temperatures within the server and may reduce fan speeds or shut down unused CPU cores to help reduce power consumption. I am all for going green, but you need to balance this power savings with a potential risk, called "change of state." Power consumption is an important topic in the data center and saving even a few watts per server over the course of time can add up to savings. However, the increased risk of failures is often overlooked. Failure to electronics and hardware has many causes, such as manufacturing defects, abuse or heat. Often times our data center equipment is left on for years without shutting it down, except for the occasional reboot. Having lived through a large scale data center outage, one of the key takeaways was the sheer amount of equipment that failed on startup. The outage was close to 40 minutes, giving the equipment that had been running at its operating temperature for many years to get cold. Unfortunately, a variety of equipment could not handle this change in temperature and failed when power was restored. Does saving a few dollars per year balance the risk of increased outages?
Where server vendors can add value
These arguments support the idea that the hardware is a commodity and it also helps to reinforce the other keys aspects of configurations, software and value-added resources. For example, selecting hardware with a choice between blades and rackmount servers has been a topic of intense debate among professionals. Each has a place and function in the data center, and if your hardware vendor has the ability to provide both platforms, it gives you internal flexibility to stay with a single vendor and minimize installation and management complexity. However, some of the other value-added services, such as engineering and architectural services, may not be best served by the hardware vendor that will not be as well-versed as organizations dedicated to this job.
The one piece that can be the deciding factor in whether a hardware vendor matters is -- oddly enough -- the software. When a server breaks, it's swapped out and replaced with a similar appliance. It's the software that makes the hardware unique. Each vendor brings different experience and features to its hardware with the software that it includes in a server for virtualization. This can include remote control, automation APIs and interfaces. The box itself is becoming more ambiguous each day. Even converged infrastructure is starting to see some of the same trends. VMware was on to something several years ago when the company introduced the concept of the software-defined data center, and this trend has no sign of slowing down.
Next-generation hardware for your virtual data center
Choosing the best hardware for virtualization
Data center hardware upgrade guide
What to consider when selecting a server for virtualization
White box servers vs. name brand servers