If you want to be the best in data center operations you are right to benchmark yourself against the cloud computing leaders – just don’t delude yourself into thinking you can match them.
In our latest research report, Rich Fichera and I updated a 2007 study that looked at what enterprise infrastructure leaders could learn from the best in the cloud and hosting market. We found that while they may have greater buying power, deeper IT R&D and huge security teams, many of their best practices apply to a standard enterprise data center – or at least part of it.
There are several key differences between you and the cloud leaders, many of which are detailed in the table below. Perhaps the starkest however is that for the clouds, they are the product. And that means they get budgetary priority and R&D attention that I&O leaders in the enterprise can only dream about.
But at the end of the day, clouds are a bunch of virtualized servers and storage racked and stacked in an air-conditioned building, just like you are. Ok, there are a few exceptions, like Microsoft’s newest Windows Azure data centers in Boydton, Virginia that are basically outdoors, but otherwise mostly the same. So what can you learn from them? Well you’ll have to read the report for all the lessons, best practices and recommendations, but as a starter, here’s some of the thinking:
The cloud (and hosting) leaders have a huge advantage in that they mostly have large pools of homogeneous, virtualized infrastructure. Most of us on the enterprise side have one of everything that’s been built by the major vendors since 1995. However, increasingly we are standardizing a part of our infrastructure – the virtualized x86 pool. And here is where cloud lessons are best applied.
1. Standardization & Simplification. While it is simple to say you should standardize the infrastructure in your virtual pool, few of us actually do this. It’s our nature to play the vendors off one another to get that extra 15 percent off the next batch of servers. And we, too often, get suckered into the latest server, converged infrastructure or unique configuration that the vendor claims will make performance go up or simplify our lives. When you vary from a standard configuration, rarely do either come true.
2. Workload Consistency. I know, I know. You’re shaking your head at this one. To think that we in I&O (the tail) can wag the dog (App Dev) and affect workload architecture is nuts. But that’s what the clouds do – and you can too. The key is in providing the right incentive. Application developers adapt to the cloud (not the other way around) because the cloud gives them what they want most – autonomous agility. And if you did the same you would find similar adaptation.
3. Automate Everything. How well does your infrastructure perform when your best administrator is out for a week’s vacation? If your stress level just went up, then you need this advice badly. Automation is the key to efficiency and meeting SLAs for the cloud leaders. They automate everything they possibly can to deliver the autonomous agility above and to drive efficiency and consistency in operations. By minimizing human actions they prevent human error and shield themselves from the rock star admin. And this frees up their operators to focus on higher value tasks such as new service delivery, infrastructure architecture improvement and capacity management.
4. Maniacal Control over Power & Cooling. When you operate at the scale that clouds do, a small rise in the cost of electricity can take ten points of margin away from the service. That’s why they are so focused on data center design, layout, cooling and geographic location. While you probably aren’t nearly as sensitive to power costs, you can reap the rewards of their efforts through what’s come out of their efforts: new open data center layouts, power efficient components and cooling technologies like airside economizers that can dramatically lower facilities costs and availability.
Whether you are in the market for more data center space or just trying to eek out more efficiency from your virtual server farm, there are lessons to be taken from those who do this the best. Even if I&O isn’t the product, that doesn’t mean you can’t treat it as such.