Just click a button and you can have a new server online running Linux, Apache, MySQL and Rails in 15 minutes!
Never, ever log into a server again – it’s all automated from patches to new code deployments!
Zero downtime – it’s all automated so it can never fail!
Which one of those pipe dreams have you heard today?
Data Center Automation – the holy grail of IT departments everywhere. It’s actually a really good idea, in fact it’s a necessary component in most organizations because of the sheer complexity of IT. Who in the world would actually want to maintain 100s or 1000s of servers the old fashioned way? Why would a company risk rolling out new deployments by having someone sit down at the keyboard and type in the commands manually to start the process? Yet for all the rewards that data center automation promises, the path to getting there is filled with some harsh realities.
Reality #1: You have to do it manually first.
You can’t automate what you don’t understand. Even I was a big proponent of this in years past. Just automate everything that doesn’t move – and if it does move, figure out how we can automate it’s movements! The reality is that you can’t automate what you don’t understand. Automation tools can help make sure you get it right every time, but you have to know how to get it right the first time.
Reality #2: It takes time. Lots of time.
Not only will the technical challenge be massive, but the cultural change will be enormous! Rome wasn’t built in a day and any company thinking they can roll out data center automation in a few weeks (or even a quarter) is dreaming. Sure, you might get lucky and be starting off with a “Greenfield” project, but the reality is we all have legacy systems and processes that we have to work around – and in many cases change. Change takes time. Doing it right takes time. Making changes so that you do it right consistently while keeping the systems up and running takes a lot of time.
Reality #3: The culture shock can kill.
Say the word automation and I guarantee you will find a dozen people in your organization within the next 5 minutes who think you mean that you are going to automate them out of a job. The cultural change in any IT organization to adopt data center automation is huge. People have to be willing to give up “doing it themselves” and instead transfer their knowledge the tool. The good thing is that once they have done this they are now freed up to work on the more important stuff. Would you rather be building your 5000th Windows 2008 server or would you rather let the tool do that for you and you work on implementing new virtualization technologies?
While it sounds like an easy question, many people equate the word automation with “job at risk”. Throw in “change” and you get complete chaos at some companies.
Reality #4: Tools won’t solve process and people problems.
Some people think they can install Puppet or Bladelogic and all their problems are solved. However, what many people find out is that their organizational problems are things no tool can fix. This relates directly back to change, but also encompasses the process by which we do our work.
Automation is not going to help your IT department if they still have to go through a dozen manual approval processes and the company isn’t willing to shift people and responsibilities around. Throwing a tool around a broken process just gives you a colossally expensive broken process!
Think DevOps!
Reality #5: It’s OK to take small steps.
You may want full end-to-end data center automation. Everything from server provisioning to compliancy checks are done automatically without anyone getting in the way. That’s a great goal – and one I think every organization should work towards. The problem is you aren’t going to achieve it the first time out of the gate.
Set realistic expectations and know that each small step is helping get you a bit closer to utopia. Don’t be afraid to follow the Agile approach of delivering things in small chunks and coming back and improving on it later.
Your end users would love single-click server deployments; but chances are for iteration #1 they will be pleased as can be if they can just get a base OS installed and running! Show your value incrementally, and the business (and your coworkers) will support your efforts enthusiastically.