Data Center Complexity – and Clarity Around Outages

Cloud outages get an inordinate amount of attention – it’s always interesting watching the tweet stream and feed reader after I hear the first inkling of a Cloud outage – all of a sudden my world is filled with people saying that this particular outage really marks the death-knell for Cloud Computing. I’m lucky, not being an employee of any Cloud vendor which allows me to take a back seat and watch the exchanges dispassionately.

It was only while researching for a CloudU report on the complexities of a modern data center, that I realized just what goes into a modern data center. From things like physical security and multiply redundant power and networks to incredibly complex networking and load balancing components – the modern data center is an incredibly complex beast.

And yet they still suffer outages.

This got me thinking and it made me realize two things.

Outages for Public Vendors Get an Order of Magnitude More Coverage

I spend a bunch of time talking to people who work within enterprise – both line of business stage and also IT folks. The great untold story inside most of those organizations is that they battle outages themselves, and much more frequently than you’d expect. I’m often seeing reports from people who can’t work because email is down, or their application server crashed or whatever.

The truth is that, were we able to compare the relative up-time from Public Cloud providers with the up-time for organizations running their own infrastructure, the results would tell a story that no end of fear-mongering can change – regardless of how many “9”s we put as a figure for up-time expectations, and no matter what the SLA says, from a reliability perspective you’re better off letting someone else have the sleepless nights.

Complexity is Increasing, Wash your Hands of It

As we detailed in our report, the complexities involved in building, maintaining and securing a modern data center are…. well, complex. I’ve spent time wandering around data centers, and spent years helping to build them a couple of decades ago. The differences in complexities in that time are stark – and if discussions with my more forward-looking colleagues are anything to go by, those complexities are going to do everything but diminish going forward.

Given the opportunity to abstract responsibility for what is, essentially, an undifferentiated service, and one which is supremely complex to boot, is something that all organizations should look at as an attractive option.

This series of posts are companion pieces to the CloudU series of educational material. We’d love you to join in some of our webinars or read the whitepapers the CloudU homepage is here, And you can register to have updates sent to your inbox (in a non-spammy way of course!) as well.

Enhanced by Zemanta


  1. Unfortunately the typical end client in the industry will never understand the complexity behind operating a data center and providing nonstop-service. They often just focus on the number of 9’s they see promised, without even considering what the providers actual setup is and how the SLA is put together. And what is a promise of five nines worth, if the SLA doesn’t back it up?

    You can design your setup as redundant as possible on all levels, but something can always go wrong – either due to failure or human mistake. And at the same time, as you make it more and more redundant, you add more and more complexity in to the setup and thereby get exposed to more potential types of failures. In some scenarios you even risk ending up having a setup that is so copmplex, that when it eventually crashes down it will take a long time to get it up and running smoothly again.


Please enter your comment!
Please enter your name here