Knowing what’s in your data center

  by Andy Ormsby | 29 Oct 2008 | Permalink |

This week’s Economist (October 23rd 2008) contains a special report on “Corporate IT”. A section of this talks about data centre and contains some great statistics. One of the more surprising to many readers may be the statement that “nearly 30% [of servers] are no longer in use at all, but no one has bothered to remove them.”

How on earth could that happen?

Computing is a funny business. Some of the things that sound hard to do end up being relatively easy, others that sound like they should be straightforward end up being horribly complicated.

One of the many things that Tideway Foundation does for our customers is tell them what servers they have in their data centres, what software is installed and running on them, and how the business services (which is, after all, what the data centre is there for) map on to that infrastructure. The first part of this, knowing what servers you have in your data center, sounds pretty trivial. Surely you can just go and count them?

Well yes, you can. The problem is that counting them doesn’t actually help you understand what the server is doing, whether it is important to the business, or indeed whether it is doing anything useful at all.

Unfortunately the data centre is full of anonymous racks of computers which all look much the same. If you can find your way into the data centre, you can see and touch the boxes, see the cables that supply them with power and those which connect them to the network, but it won’t help you understand whether the server you are looking at is running part of the company’s latest and greatest Web 3.0 social networking project or the remnants of a long abandoned CRM system.

One way of thinking about this problem is to understand it as the divide between a physical view of the infrastructure, stuff you can see and touch, versus the logical view, which involves understanding what software is running where and how it interacts with other bits of software running elsewhere.

A systems administrator can gain a logical view of a server by logging in to it over the network and executing commands which will tell him or her what software is running at the time. But repeating this process several thousand times to build a complete picture of the data centre simply takes too long. By the time the systems administrator has finished, the world has changed. And the resulting view is likely to be limited to those servers that you knew existed. How does the systems administrator know to log in to a server if nobody knows it exists? This is more more likely than you might think since a typical systems administrator in a large organisation may never even have visited the data centre.

How does Tideway Foundation help?

We automate much of this process, providing a detailed view of what servers are present, how they are configured, the software that runs on them, and the business services they support. It might sound easy, but is harder than it looks.

One challenge is that most data centres have accreted many layers of technology and complexity over the years, so servers are often of many types with many different operating systems and many different versions of those operating systems. A complete picture may require hundreds of distinct software products and versions to be identified. Being able to cope with this heterogeneity is complicated, but essential to provide a complete picture.

The good news is that we regularly find servers that are not being used or which can be retired for other reasons. Given that it can cost $10K-$20K per year to run each server in the data centre, and that large data centres may contain many thousands, the potential for cost saving is huge, even if you don’t find the 30% of unused servers that the Economist article suggests may be present.