Data center cost varies from year to year depending on changes in volume, technology, customer need and what service the data center is offering. We continously struggle with cost; work smarter, buy cheaper, utilize the equipment better. But there is also an underlying development of cost depending on how data center costs are allocated.
We all know that price / performance when it comes to hardware is getting better and has been since the early days of computing. This means that hardware cost for same capacity continously is getting lower. People, premises and nowadays also energy are costs that are rising. With a fixed number of people and no change in the energy consumption a rise in the overall cost of the data center can be expected. Different types of cost exhibit different kinds of growth and depending on the cost allocation between these types, data center competitiveness over time will be affected.
Traditionally, data center cost has been divided into 5 categories; hardware, software, people, premises and other. The increase in cost of energy, as well as from a political point of view, has made it necessary to create a sixth category by lifting energy from premises. These six cost categories have different characteristics both in terms of cost development and how they can be affected. For the energy category it is even reasonable to speculate that sufficient energy is not available.
Let us look at a simple example: Hardware cost is decreasing with 15% per year and the cost of software, people, energy, premises and other has a total increase of 8% per year. In the following diagram we see the change of cost over time as a function of hardware cost share in total cost.
In the example, with a yearly decrease of hardware cost of 15% and a yearly average increase of 8% for other costs, a data center with less than 35% hardware cost will have an automatic total cost increase. With 60% hardware cost the effect is a decrease of about 6%.
As shown here a data center starting at 60% hardware cost share has an underlying competitive edge over three years of 26% compared to a data center starting at 20% hardware cost share. Often this fact is hidden by increases in volume and the introduction of new technologies which makes direct comparisons difficult.
Is this discussion really valid as server farms volumes and technology quickly changes?
If we make a historical comparison with mainframe then the answer is yes. Server farms will just like mainframes develop to become a mature and stable market. Virtualization and subsequent technologies rapidly change the server farms and the market will be mature within a few years.
To further show the value of studying the underlying cost development we will take a look at energy. Assume that energy costs are going up with 10% per year and that all other costs are going down with 5%. This will produce the following diagram:
If energy costs are more than 30% of total cost we have an underlying cost increase given that the hardware cost reduction is large enough to produce an overall cost decrease with all other costs of 5%. We can also tell from the example that over a three year period a data center with a 5% energy cost has a competitive edge of 12% over a data center with a 30% energy cost share at the start of the 3 year period.
What are the advantages of studying the underlying cost development?
- A strong or weak underlying cost development can be hidden by a snapshot study of total cost in relation to data center size. To show if a data center has a positive underlying competitive edge or will require improvements just in order to stay at current levels could prove to be very valuable information.
- Change of technology and or methodology can be described by how the split between different cost categories is changing and with that the data center cost development, all other factors excluded.
How this discussion affects partial outsourcing is discussed in my blog "Data Center Cost Allocation and Effects on Partial Outsourcing".
