The use of virtualization is pretty well accepted these days and you find it pretty much everywhere, from the mid-range consumer PC to the high-end enterprises Virtualization is getting lots of play.
Using virtualization 10 or 12 years ago, yes that long ago, was almost bleeding edge and forget even suggesting running it in production. Most folks saw virtualization technologies as a nice to have not a need to have. More recently though data center managers started trying to squeeze ever more utilization out of their existing environments through hardware consolidation and scale-up efforts only to find that this was no magic bullet and required something to take the pain out of these massive projects and increase success rate of consolidation and scale-up/out implementation. Virtualization was indeed an answer but many who had considered it almost a toy were still unsure if it could handle the stress and load of “production” capacity and so only virtualized items they felt were less critical. This is where the recent tidal surge of virtualization has come from inducing a subsequent wave of “Cloud” technology hype.
Where some see a challenge others don’t see any. The use of virtualization is a huge boon to companies trying to squeeze every last penny out of their investments and long term costs but it requires care and feeding in a somewhat different way than non-virtualized infrastructure. The main reason is mathematics. When consolidating and a solution is sized properly, servers wind up being put onto a single pool of resources as virtual machines, or VMs, which require either the same or slightly less resources than their physical counterparts because all resources are used as a pool. The math problem comes from something call “over subscription” whereby a data center manager or administration team will intentionally add “too many” hosts for the available resources but still create the VMs with the “proper” resource sizing known to be needed in the physical world. This is basically a bet that not all the resources on a given host will need to be used at the same time and thus there will always be free resources available. Another part of the math is simply the number of servers now placed on the VM Host.
In a given scenario where 100 physical servers are consolidated down to 2 or 4 VM Hosts, those servers, if not reduced in total number from 100 will all be housed on the VM Host, but then additional services may be needed and result in new servers being created or “spun up” as we say in the biz. With over subscription this activity is almost limitless without safe-guards put in place. (like using a hierarchy of resource pools and assigning rules against them to stem the tide of over-subscription) Due to the capability of being able to oversubscribe though, data centers have a very rich, organically grown sprawl of virtual machines that ultimately increases the burden on IT staff to support such infrastructures.
Without seeing this challenge, you just sprung a trap on yourself. This is a common scenario in which folks who begin to enjoy the ease of use and increased capability find themselves. Servers, everywhere. This means that once servers go from physical to virtual that the DR, backup, and ultimately recovery and management scenarios must change as dramatically as the infrastructure.
Another challenge that arises from all this consolidation and capability is more of a people problem. The problem of job displacement. Responsibilities that were once tied to a specific group or person within the IT staff now becomes more of a shared responsibility. The environment is sprawling with virtual server fauna if you will and everyone who works in the environment has to know more about what’s going on with each person involved in IT. Once this happens though and you now have a broad set of resources with which to manage the everything, the next step is often reduction of those resources because IT staff has become commoditized to a degree, depending on the level of specialization within the infrastructure of course.
Adopting this paradigm is a huge mistake.
The resources you have before you virtualize can be readily reassigned to support individual or groups of business units, improve service times, increase profits, prevent outages, and many other crucial IT related functions. What often occurs though is the numbers look so good after consolidating is that someone in upper management often suggests saving “just a little more capital or OpEx” cost by headcount reduction. It’s a short-term win but a long-term loss since it is inevitable that the good ones will leave, and the mediocre will stay, and without significant improvement in human capital (i.e. training) businesses will often suffer from brain drain and their new found freedom, flexibility, and competitive edge will have been reduced significantly. Math strikes again.
Cloud … Could I?
Profit center, cost center, how about competitive center? Many companies have developed a way of utilizing virtualization as more than a means to an end, but rather an on-going process to accomplish many different kinds of goals and employ more than one kind of virtualization for the right reason to avoid putting all their eggs in one basket as a company but also to remain nimble in the face of an ever changing landscape. The hype calls it Cloud, I call it ubiquity. Omnipresent computing infrastructure that never goes down, never breaks, and never needs maintenance at a general level which also *never* shows signs of aging, bottlenecks, has unlimited capability for resiliency and scalability. That is this so called “cloud.” It doesn’t exist, anywhere, but we do have parts of it and someday soon, we might actually see a system which accomplishes this Omnipresent computing. One key component of that infrastructure is Automation. Virtualization infrastructures allow for this type of automation but on the small scale without lots of human capital to manage it. Once *simple* automatic deployment and provisioning capabilities are widely available and *very easily* integratable regardless of the virtualization infrastructure, then we may get closer to this “Cloud.”
You can use virtualization today to garner in a new era of computing and that is the virtualization strategy you need to adopt in order to provide the level of service the needs of each business unit. First work on the Physical-to-Virtual migration and decide what services will get consolidated down to which virtual machines and keep the relative number of machines low but scale up their virtual beefiness. This will allow a greater degree of flexibility without the need for massive recovery infrastructure jobs, etc. Keeping the relative number of machines low will improve responsiveness to applications which reside on them as well as provide simplicity in management and allow easy use of technologies such as snapshots.
Next prepare for scalability by adopting a lego-block type of approach when budgeting for and building out the infrastructure. This will help to maintain the current resource utilization curve the environment has as well as keep performance in check since bottlenecks are well known after the P2V migration exercise you just performed. Use 10Gbps network connections where possible and try to keep the hardware footprint to a minimum to avoid large amounts of capital asset depreciation.
(i.e. servers are commodity usually if built small enough, storage might even be the same way depending on the contents of your “lego-block.”)
The Kitchen Sink
Don’t forget Backup, DR, and remote access capabilities and technologies and process updates. As I already outlined, these are usually crucial and if you downsize physical servers and increase your capability you will rely on Backup, Recovery, DR and Remote Access a ton more than you ever did before. No consoles to log on, just remote terminals. Might as well find a way to add this to your arsenal of company competition capabilities. (say that five times fast)
Don’t forget the user-base. With all these changes they might start feeling like you don’t care about their needs and complain that “IT doesn’t listen to us” or that “IT is so stupid. They just took X server offline and that’s where Y app was” etc. User acceptance testing , UAT, is often critical in determining how successful IT is performing for business related goals, especially when you’re talking about a large effort like consolidation. Users are fickle and finicky, but if you listen to and include them in large transitions like this they might surprise you by actually helping IT efforts when virtualizing, consolidating, and recovering.
Don’t forget where your headed. Keeping to the plan once it is in progress is important and anyone will tell you that scope-creep is a huge problem, but so is simply losing focus on the end goal:
Transforming your existing environment from a cost center to a competitive center.
Keeping your focus and communicating through the transition as well as your long-term goals and strategy as a business unit like all the rest will help you get there, save the day once in awhile, and allow you to increase budget. This is of course so long as you can show how the strategy allows the business to make more profit. Doing this will make everyone happy, and keep everyone else off your tail.
Don’t forget the desktops either! Virtualizing desktops is a great win for a company that wants to virtualize infrastructure but has to be done as a discreet process either before or after the major infrastructure changes for two reasons. VDI as it is know, can either cause the changes or be a product of the changes I’m writing about. How you proceed into that area is up to you but DR will get a whole lot easier if you do go down the VDI path.
We’re not talking keeping up with the Jones’s, we’re talking about going someplace the Jones’s aren’t even going. IT that helps the company make money because it is truly integrated into every facet of what the company is doing, where it makes sense of course, and introducing a level of resiliency, flexibility, and nimbleness that leaves a company moving ever forward and not stagnant and aging. It does mean a lot of changes, but it can be done, it takes time, effort, a lot of effort, even more effort, money, some more effort, a huge amount of communication, consulting time, buy-in, dedication and commitment, and good people to manage all those changes and provide the business with a good comfort level.
Did I mention effort? It can take months to roll out a highly virtualized environment in a large enterprise, so it should be looked at as a process, but just because it takes less time in a smaller environment doesn’t mean the same methodology shouldn’t apply. It does.
Plan, prepare, deploy, manage, upgrade, scale and repeat.
After all this one day you’ll ask yourself why you didn’t virtualize sooner.