CIOs Need To Realize That Virtualization Isn’t All That It’s Cracked Up To Be

by drjim on August 3, 2011

Sure Virtualization Seems Neat In The Beginning, But…

Sure Virtualization Seems Neat In The Beginning, But…

To read the IT trade journals or speak with CIOs you’d think that we’ve all found the magic silver bullet that IT’s been looking for during the past few years: server virtualization. The ability to mash together a bunch of different expensive individual servers and shrink the company’s IT footprint down by a factor of 5x while reducing power and cooling costs at the same time sure seems to be a miracle cure for IT budget problems. Guess what: this isn’t Hogwarts and you’re not Harry Potter. Virtualization has its own set of problems and we need to have a talk…

What Is Virtualization?

So first off, let’s make sure that we’re all on the same page here with our understanding of just exactly what this virtualization thing is. In the past, IT departments used to set up a new server for each new application that they wanted to deploy. This resulted in the IT department having to maintain farms of servers that were all horribly underutilized.

The arrival of virtualization software changed everything. This low level software allowed multiple applications to run on the same physical hardware but believe that they had the box all to themselves. Now you could combine multiple individual servers into a single physical box. Things like what operating system an application used no longer mattered – you could mix and match to your heart’s content.

Problem: Virtual Machine Sprawl

Evangelos Kotsovinos has taken a close look at just exactly what it means to introduce lots of virtual machines into a company’s IT infrastructure. What he’s found is that although CIOs might think that this changes everything, it doesn’t.

It turns out that managing a virtual machine (VM) takes roughly the same amount of effort that managing a real box does. When you couple this with the fact that it has become so easy to set up new VMs, what you’re seeing is unconstrained virtual machine sprawl.

IT departments are struggling to keep up with more and more VMs as staff set them up and then forget about them. Every IT department now needs to come up with a VM reclamation solution.

Problem: Scaling

The very newness of VMs is causing IT departments to encounter a whole new set of management headaches. In the old days, IT departments had developed the tools and processes that they needed in order to deal with building large groups of new servers or handling a planned data center maintenance activity.

The arrival of VMs has upset this carefully established way of doing things. The problem is that often the VM management tools aren’t able to scale up to the size of enterprise operations. This leaves IT departments struggling to find ways to manage the beast that they have created.

Problem: Troubleshooting

There’s something deeply satisfying about tackling a system problem when you have the physical box in front of you. You know that you can always reach out and swap out various components if you have to. The same is not true when you’ve virtualized all of your servers.

Kotsovinos points out that a VM is really a collection of interconnected physical subsystems: server, storage, and network. When you are dealing with a system problem, like a slowdown, it’s going to require a whole new set of skills to track down what’s really going on. Additionally, virtualization is so new that often the right tools to do this type of trouble shooting may not exist yet.

Problem: Silos

Think about how your IT department is set up today. Generally we draw lines between various disciplines based on what they do: the Unix team, the Windows team, the storage guys, the network guys, etc. The arrival of virtualization in the data center is going to screw all of this up.

The reason that virtualization can cause such a disruption is because issues that have to do with the VMs more often than not involve all of the various disciplines. No longer will the storage team be able to just focus on storage issues. Instead, they are going to have to work together with several other teams in order to try to solve complex problems.

What All Of This Means For You

Server virtualization is a fantastic discovery. However, CIOs need to realize that it’s not going to make all of their problems go away.

Instead, virtualization is going to end up replacing one set of problems with another. These will include potentially unchecked virtual machine sprawl, scaling issues, more challenging troubleshooting, and a breakdown in the IT silo structure.

Face it, virtualization is going to take over both the IT back office and probably the IT front office eventually. CIOs need to understand that as this occurs, we’re all going to have to adjust how the IT department is run in order to meet the new set of demands that virtualization is going to put on us…

– Dr. Jim Anderson
Blue Elephant Consulting –
Your Source For Real World IT Department Leadership Skills™

Question For You: What do you think is the best way to keep virtual machine sprawl from getting out of hand?

Click here to get automatic updates when The Accidental Successful CIO Blog is updated.
P.S.: Free subscriptions to The Accidental Successful CIO Newsletter are now available. Learn what you need to know to do the job. Subscribe now: Click Here!

What We’ll Be Talking About Next Time

Let’s face it, CIOs just like everyone else are currently being overwhelmed with too much information on a daily basis. Things were bad a few years ago, but with the arrival of blogs, wikis, Internet video and smartphones, there just doesn’t seem to be any way for us to keep up. That’s why more and more CIOs are turning to simulations in order to visually present large amounts of information in a way that we can absorb it. However, it turns out that our simulations just might be lying to us…

Be Sociable, Share!

{ 2 comments… read them below or add one }

Michel Labelle August 3, 2011 at 5:49 pm

The issues you describe are no different than the traditional 1-application 1-server approach that we have seen for years in the datacenter. The key is to use virtualization as it is supposed to be used, as an element in overall data center and application consolidation.

Here is a simple 4 step process to getting your data center costs under-control today.

The first step to success, is consolidation of applications. Put more on an existing box. This achieves the highest savings as you physically remove boxes, power AND management. Expect savings directly proportional to the number of servers you are removing. Note, you don’t need to replace the existing servers you have in place today, simply consolidate application workloads on the SAME existing servers you have today. Easy 20-30% savings (capital and operating) are typical. Change management stops application and box creep dead as no one will be able to justify putting an existing application at risk simply to install a new one. By consolidating applications to a departmental server, you also increase the support for DR, future upgrades and allow the business to justify the needs vs. IT dragging the business forward.

2nd step is consolidation of boxes. Put in blades or other high density technologies to reduce the physical foot print of your data center further. When it is time to replace, now start looking at reducing the physical demands. This is not to avoid direct server capital costs, but to reduce physical data center growth costs. Expect being able to defer data center infrastructure (power, cooling and physical space) costs by 3-5 years over traditional build outs. This is how IT can “save” money by showing cost avoidance. Entirely Capital savings. Figure 25% if you can defer upgrades by 3 years.

3rd step is virtualization of servers and workloads. Even with re-using existing servers and collapsing many application workloads on a single server (step 1), there are still great reasons to deploy virtualization including DR, HA and remote (lights out) management. This allows data centers to be placed further away from the operational environment (co-location, remote hosting etc). Additional savings in terms of licensing, test/dev and DR hosting. Expect an additional 10-20% operaitonal cost savings.

4th step is cloud (private/public/hybrid) with charge back. If you are running your IT business right, you know IT never installs a server because it wants to. Having an effective charge back mechanism that can span private/public clouds will give the business a true cost allocation and allow them to make informed decisions. It is too easy to justify a $10-50k capital box as part of a project. By placing a full cost accounting model around the application, hosting and support of that application you get a lifecycle cost that keeps the demands for new servers in check. Entirely operational costs. Figure 10-15% initially and 25% over the lifetime of the system.

The key is to remember that a server, be it VM or physical is NEVER free.


Dr. Jim Anderson August 5, 2011 at 8:42 am

Michel: Great points! One impact of all of this server consolidation is that all of a sudden your disaster recovery plan gets a lot more important. No longer do you have to worry about just losing one server / one application. Now if you lose a server multiple applications can be impacted. It is possible to deal with this, but you need to have a plan…!


Leave a Comment

Previous post:

Next post: