Planning and designing your IT Infrastructure – Part II: Evaluating your requirements and designing your data center

In my last blog post, i talked about all these little things you need to keep in mind when setting up a new or replace your current network infrastructure. I came up with that topic first, because in our case we are not only building a new data center, but we are also building a whole new office location. Providing a working network infrastructure to the office floor was crucial in our case, that’s why we took care of that first.
In this blog post, i will talk about planning a data center itself. The most important step in data center design, is to evaluate what you want and what you really need: Pretty much everybody wants a Tier 3 design, but hardly anyone needs it. In fact, with todays supply of professional webhosting solutions, colocation centers, cloud computing and other solutions, you need to evaluate if your company really needs to build their own data center at all. Which brings me back to remind you of the 2 most important rules when planning it infrastructure: First of all: Don’t overbuild, it’s a waste of money! Second of all: Don’t plan for now, plan for the future! Sounds contradictory? Well, it really isn’t… 😉
Don’t overbuild!
Before you go ahead trying to find a location for your new data center, calculate your UPS supply, thinking about air conditioning, air flow management, redundancy, raised floors, server racks, hardware, storage, networking and what not, find out what you really need. Helpful information can be found with the Uptime Institutes tier level concept, which uses a standardized methodology to determine availability of the physical topology in a data center [1] [2]. There are currently four levels:

  • Tier I: Basic non-redundant site infrastructure. Non-redundant capacity components. Expected availability: 99.67%
  • Tier II: Redundant capacity components site infrastructure. Meets Tier I requirements. Provides redundant infrastructure components. Expected availability: 99.74%
  • Tier III: Concurrently maintainable site infrastructure. Meets Tier II requirements. Multiple independent distribution paths service the IT equipment. All equipment must be dual-powered. Expected Availability: 99.98%
  • Tier IV: Fault tolerant site infrastructure. Meets Tier III requirements. All cooling equipment is independently dual-powered. Fault-tolerant electrical power storage and distribution. Expected availability of 99.995%

The difference between the tier availability seems next-to-nothing, but could be significant depending on the application: Tier I allows ~1729 minutes of down time, whilst Tier IV would allow only 26 Minutes to meet the tier certification criteria. Even though it’s hard enough to meet these criterias, it still should give you a general idea of what you are planning for: Determine your data center business goals, the possible/acceptable risk profile and the available capital. This sets the foundation for a proper data center design.
Don’t plan for now, plan for the future!
That would be the next step to consider. As i said in my last blog entry: You probably want to add some extra resources to easily be able to extend your infrastructure at a later time. Make sure to have all information necessary to properly design your future data center. What are you building the data center for? What kind of growth does your company expect in the future? Do you want to add the possibility to re-purpose your data center at any later point in time? Keep all possibilities in mind when adding additional resources that might be useful in the future.
Cost of Ownership!
Always take your total cost of ownership (TCO) into account! It’s easy to get yourself trapped in focusing on initial cost only! It might not seem that expensive to build a data center (well, of course we could have a philosophical discussion about defining expensive here, but lets keep that for later). However, maintaining a data center can be staggering. There are three basic TCO parameters:

  • Capital expense: Basically building the whole thing – Not only the construction costs itself, but also all costs for planning and designing are added here.
  • Operations and maintenance expense: Maintenance costs are the costs associated with proper maintenance of all infrastructure, for instance license costs, maintenance contracts or soft/hardware upgrades. Operating costs are the costs associated with the daily operation, for instance personnel, training and procedures.
  • Energy costs: It is essential to calculate any possible future development in energy costs. Electricity prices will be a MAJOR parameter in the future, don’t get yourself trapped in underestimating future costs! Also, this is where you might want to think about investing extra capital to optimize energy efficiency in your data center: A growing number of data centers are redirecting heat from their servers to air condition systems to heat their offices or support their hot water supply. There are lots of possibilities in how to use your data center to lower energy costs at your other locations [3].

Select a proper data center site!
A common mistake is to start searching for what you think would be the perfect space to build a data center, before having your data center design in place. Without having a proper data center design, there is not much sense in spending time looking for a possible location. The main problem with selecting a site before even having set up your design requirements, is that the site possibly can not meet these later determined design requirements! While it surely is convenient to have your data center located in a basement of your office building, it might turn out to be extremely counterproductive if your data center is located right next to your buildings main water supply facilities. Depending on the desired tier and/or security level, you have to take into consideration power availability (which, for instance, also includes diesel supplies and contractors) , connectivity, accessibility and other possible geographic issues such as earth quakes, flooding, tornados and other disasters. [4]
Once you answered all the above questions, your next step will be to take care of all the “little” details such as thinking about your possible server hardware, calculating power efficiency, choosing a UPS solution, thinking about air condition and about a thousand other things. I will try cover these topics step by step in the next couple of weeks. Until then, feel free to comment, add any ideas of your own and point out any mistakes i made.
[1] http://www.fibertown.com/pdf/uptimeinstitutetiersystem.pdf
[2] http://www.gpxglobal.net/wp-content/uploads/2012/10/TIERSTANDARD_Topology_120801.pdf
[3] http://www.datacenterknowledge.com/archives/2011/03/18/energy-efficiency-guide-heat-recycling/
[4] http://www.datacenterknowledge.com/archives/2013/03/13/dont-delay-on-lessons-learned-from-sandy/