There's no way to avoid it: The total cost of ownership (TCO) for data centers gives every owner a case of sticker shock. Not only is the capital expense (CAPEX) -- or the cost to design, build, test and commission -- increasing, but the inevitable operating expense (OPEX) increases on the horizon are daunting. Knowledgeable users have forecasted that the OPEX will exceed the CAPEX.
Syska believes the following are the basic principles of reducing TCO and constitute best practices:
1. Spread redundancy across multiple, lower-tier data centers
Consider whether upgrading the IT network can reduce the facility availability required. For example, can three less costly Tier 2 facilities serve your needs better than two more costly Tier 4 facilities? (Tier rankings are based on The Uptime Institute's tier whitepaper).
The cost centers in higher availability design are:
- Redundant systems with redundant components in each system
- The increased operating cost that results from reduced efficiency due to a redundant system and components.
Two Tier 4 facilities with a four-megawatt computer load may cost $100 million each for a total first cost of $200 million, while three Tier 2 facilities of four-megawatt computer load may each cost $46 million for a total first cost of $138 million. The difference in first cost may be $62 million. The tradeoffs have to do with network costs versus facility costs. Network costs will be higher for three facilities than for two.
2. Base data center site selection on utility costs
Consider the present and estimated future costs of utilities in site selection. Sustainability must be considered in every project.
The best sources of information on utility rates for data centers include:
- Knowledgeable national specialists such as CB Richard Ellis, Jones Lang LaSalle, Julian Studley and others
- The utility companies' published rate schedules and in-house experts
- Watching where others are going. In my recent experience, a low electric power rate is $0.038-$0.04/KWH; average is $0.08-$0.12/KWH; and high is $0.25/KWH.
3. Consider OPEX as well as CAPEX
Focus on long-term operating costs (OPEX) in addition to initial costs (CAPEX). For example, consider a site that has, at present, low electric utility costs and medium water utility costs. Air-cooled water chillers will provide the lowest initial cost, but rapidly escalating electric utility costs may prove that water-cooled water chillers are the better choice. Energy and financial modeling based on estimates of future utility costs will reveal the best decision.
Most knowledgeable owners use a TCO model that considers both CAPEX, OPEX, cost of funds, desired return on investment (ROI) and other factors.
4. Focus on scalability and modularity
If the initial UPS need is 50 watts/square foot and the ultimate need is 150 watts, design for 75 watts initially that can be later upgraded to 150 watts. If the initial raised floor need is 20,000 square feet and the ultimate is 40,000 square feet, design for 30,000 square feet initially with an additional 10,000 square feet shelled in for future fit out. When faced with a choice between quality and quantity, opt for quality that can be upgraded in the future to provide added quantity.
I try to avoid hard and fast rules since each client's needs are different and worthy of independent thought. Some upgrades are simple to accomplish and suggest ratios for consideration. For example, on Day 1 we install two UPS systems on capacity N to supply N computer load in a system + system arrangement. On Day 2 when the owner needs 2N computer load, we can add another pair of systems or we can add a third system of capacity N to the two original systems and reconfigure the arrangement as 3N/2.
5. Keep data center design simple
Keep the design as simple as possible. Simplicity eliminates unnecessary components, and cost increases reliability and maintainability.
Here are some examples:
- Minimize the number of layers in the electrical distribution system to reduce cost and make selective coordination of the electrical system easier.
- Eliminate secondary chilled water pumps by making the primary chilled water pumps variable speed.
- If all of the computer equipment is dual-cord and there are no motor loads supplied by the UPS, then static transfer switches downstream of the UPS systems are unnecessary.
6. Avoid one-of-a-kind or serial number 1 products
Use standard components and ratings wherever possible. Serial number 1 always has a high initial and replacement costs. As an example, 4000-ampere switchgear bus is more cost effective than 5000-ampere bus.
I recently heard from a client that has 12-year-old UPS equipment that is no longer manufactured has some custom, one-of-a-kind control boards. As UPS equipment ages, the availability of spare parts and trained technicians decreases.
Additionally, automate normal and emergency facility operations as much as possible. Lastly, utilize technology instead of people wherever practical.
Follow these practices and you will be on your way to designing a cost-effective data center.