Warning: This is a Touchy Topic!
Bad (storage, server, and network) configs are a few examples of why virtualization fails but these are technical in nature and are often easy to fix – whereas what I’m about to cover goes deeper.
They are the problems talked about behind closed doors.
Let’s talk openly about the real reasons why your virtualization efforts are failing…
Why Virtualization Fails
1. Cowboys Make Good Movie Characters but Hurt Infrastructure Teams
In the movie Die Hard, Bruce Willis is the Lone-Ranger saving the world; however, cowboys on an infrastructure team disrupt plans, road maps, and team dynamics.
And, left unchecked, can even be a leading reason why virtualization fails or at least why some virtualization teams don’t succeed.
Here’s an example of what a cowboy does.
Tom loves virtualization and has just attended a VMUG where he’s learned a bunch of new cool things about best practices for virtual networking. He can’t wait to get to work on Monday and implement these new ideas.
On Monday Tom arrives early and starts making changes to the production environment.
[In Tom’s defense, these changes are an improvement over the old way] but for some reason the changes have triggered alerts to be emailed to the NOC and his teammates. Something has caused the production network to go offline (not good).
Tom’s moment of gratification is short lived because unexpectedly the production cluster has gone into isolation mode. Immediately Tom jumps on vCenter and undoes his changes and 5 minutes later the alerts stop.
David, Tom’s manager has just gotten off a conference call with his boss Sarah and the VP of Technology. They’ve requested a postmortem to figure out what has happened. Guess whose user name comes up in the logs for making changes?
Tom’s actions have caused a production outage that has cost the company $500,000 in sales.
Cowboys have good intentions (most of the time) but because they tend to act independently without discussing their ideas with the team, they pose unpredictable risks.
As I said, much of the time they have the best of intentions but their actions cause team frustration, conflict, and outages.
Unless you are a one-person shop, it is best to reign in Cowboys.
2. Virtualization Free-For-All (chaos)
Because virtualization has been the rage over the last five or so years, it’s given technology teams time to go from fear to familiarity.
And because virtualization is a convergence of storage, network, servers, and operating systems, there is now a debate about who should own managing virtual infrastructure.
A mistake that can cause your virtualization to fail is allowing a free-for-all to happen (or continue) where each technology group owns and builds its own vSphere environments independently of each other.
A best practice I’ve learned is to form a team who is responsible for guiding and vetting infrastructure designs and solutions.
This team will include members of other technology groups; however, overall ownership, building and supporting of the virtual infrastructure will be the responsibility of a single team of qualified virtualization engineers and administrators.
Failure to reign in the free-for-all and centralize virtualization ownership and management will put your infrastructure and business at risk, not to mention the chaos and conflict it will create when systems fail and the blame falls on the wrong group.
3. Unrealistic Expectations to Virtualize Everything
VMware vSphere is awesome when properly architected and installed. But, this is only the beginning.
Next comes back-to-back (or even parallel) projects to standardize, consolidated, and automate every piece of hardware in the data center.
OK, I’m going to say it out loud!
Unrealistic Expectations and a lack of experience and expertise are at the root of why virtualization fails and VMware gets a bad rap…
I’ve been involved in several back-to-back data center projects that have lasted over a year each. They are great for job security but can wear out the best people you have, especially if the same people are handling all the day-to-day support, break-fix, and additional capacity needs.
Unless your environment is well architected (and this includes your software) sorting out all the dependencies on databases, apps, and web servers are going to take a minimum of 3 times longer than you think.
Any project manager who doesn’t know the environment he’s working in and says he can get it done faster is just being optimistic and will soon learn the hard way.
It’s just a bunch of P2Ving and that’s easy, right? True, however the hard stuff comes later when solving performance issues or problems that happen because networks or database dependencies were missed.
This usually happens when apps, web servers, and databases are spread across multiple locations and the goal is to centralize.
I realize most managers will do what they are asked. And most infrastructure teams are always ready to virtualize everything. However, once the unrealistic expectations for deadlines start slipping because of difficulties with P2Vs, app issues, and even people – you’ll wish you had 3 times more hours in a day…
Failing to set realistic expectations is a major player in why virtualization fails. Don’t fool yourself by thinking it’s easy.
Even when done properly there is planning, testing, trying, failing, retesting, retrying, failing again, and eventually you succeed.
Conclusion:
There, that wasn’t too painful.
Let’s summarize what we’ve learned:
- First, we have the cowboy who makes changes without properly discussing them with the team. This causes team frustration and can even cause outages. Cowboys need to be reigned in.
- Then we have the free-for-all with everyone building their own virtual environments because nowadays everyone is a virtualization expert. Centralized ownership is critical to virtualization success.
- And finally, we have unrealistic expectations to virtualize everything without allowing enough time to test and work through problems. In the end, it will take 5 times longer because of all the rework, or in worst cases, the project gets parked or canceled. Multiply by 3…
Your turn:
Please share your ideas on why virtualization fails?
