Data Center By Surprise - Data Center Cost of Ownership and Budget Planning
I don't have an easy answer to the "data center by surprise' problem and unfortunately I believe this is true for most organizations. In lieu of having all the budget & resource utilization information required to get a complete picture of the cost of our data centers we've started using a very simple model that gets us close.
For every dollar you spend on IT equipment, you eventually spend the same on the supporting facility.
Seems simple, it doesn't take your PUE into account or your resource management efficiencies, but at least it can help you get a grip on your budget planning relative to DC capacity. The understanding of cost, even at this simple level is critical to organizations developing a strategic approach to planning their data center capacity. Historically most companies have planned their data centers by "surprise". Why do we get in this trap? We know we're buying IT equipment, we have the budget numbers, in most cases we can maintain a rough idea of IT equipment spend over a relatively long term horizon (2-3 years). Why then don't we use this understanding of spend to plan for data center capacity. In all too many of our companies we seem to reach a point where the limit of our capacity is just a few months away and we're just starting to develop a plan and pull together a strategy (including budget) for acquiring more space (co-lo or owned).
The correct answer is we need to get to where we're effectively using the "Stack" to understand our data center costs.
So if we agree that there's no reason why we can't plan our data center capacity in the same way we plan our IT hardware purchases, why aren't we doing it? I don't know, is my unfortunate answer. Don't get me wrong I know some organizations have solved this problem, but for the most part I think the companies that have are also the companies who live and breathe through their data centers (Google, Yahoo, Microsoft, etc). By the very nature of the business of these large internet facing companies they have no choice anymore than a manufacturer has to plan for building more manufacturing capacity.
I don't expect to answer this difficult question in one blog or even answer it at all, but I will offer some suggestions.
We need to develop a better capacity management framework that can be aligned with corporate growth and budget planning. The data center cost for even a non-IT company of 10,000 employee's can be upwards of 20 million (one time build costs) and yet there generally isn't an owner for this cost. How can that be justified? Can you imagine building 20 million dollars in manufacturing capacity and not having a "GM" in charge of the resource? Unfortunately there are many companies spending 10s or even 100s of millions of dollars on data center capacity and they don't have a specific role identified as the owner for Data Centers. Keep in mind that I make no distinction between engineering “lab” space and “IT” data center space.
There are some options for helping to make this issue less of a problem for the C level staff. We need to improve how we provision more space. As long as we continue to provision space with 3 - 5 years of capacity our C-level folks are going to be very reluctant to figure out how to build the cost into yearly budgets, it’s just too much money. We need to further develop our ability to build incrementally or modularly. Building modularly should also give you the benefit of quicker implementation of new capacity and improved efficiency in existing space (higher PUE).
I would love to get comments on this article so I can determine whether this problem is as wide spread as I believe it to be. If you've solved this problem in your company, please share with the rest of us.