Designing a Data Center for the 21st Century
November 1999, Power Quality Assurance Magazine
Ron Hughes is the Principal in charge of the California Data Center Design Group, founder of the Sacramento Chapter of the 7x24 Exchange and a frequent speaker at National 7x24 Exchange conferences.
Designing a data center with the flexibility to function into the 21st century is a considerable challenge. With ever changing power and cooling densities, constantly changing communication requirements and with mergers and consolidations becoming commonplace, the ability to adapt a facility to accommodate different corporate uses becomes critical. The key is to build in flexibility in the original design.
There is no one design that works for every user. A cookie cutter approach to data center design is not in the best interest of the user. There are hundreds of decisions that have to be made that will impact the functionality of the data center. No two users will have exactly the same priorities. If you look at the data centers being built today, there are a number of different designs. There are however, common design criteria that most of the facilities share.
Among the design criteria that I consider the most important are flexibility to handle the changes in data processing technology, redundancy, expandability and the ability to perform concurrent operations and maintenance. Every critical piece of the infrastructure has to be serviceable without impacting the data center. The design should have no single points of failure and should include sufficient levels of redundancy to allow for both single and multiple failures. This requires the coordination of the electrical and mechanical designs.
Electrically, most data centers built today are designed so that the computer room is always on UPS power. Gone are the days of putting the data center directly on utility or generator power when performing maintenance on the UPS system. Isolated redundant, Load Buss sync, dual power path UPS designs are all different ways of achieving similar results. The relative merits of each are debatable and some are more costly than others, but the goal is the same. To design an electrical distribution system that protects the computer room from power disruptions at all times.
All parts of the electrical distribution system have to be designed with bypasses or with the redundancy that allows individual components to be shut down for maintenance or repair. Any part of the electrical distribution system that can't be maintained with the data center on line, can't be maintained, period. Most critical data centers never shut down!
The design capacity of the electrical and mechanical systems is a subject of much debate.
Most data centers are experiencing a reduction in the watts per square foot loading. Some are experiencing large increases. Again, designing in flexibility to respond to changes in technology is the best approach. Spending some additional money up front to put the infrastructure in place for additional air handlers, PDU's, chillers and UPS modules is prudent planning.
Mechanically, most modern data centers are designed with sufficient redundancy to prevent the failure of a single component from taking the data center off line. Whether designed with chilled water systems, direct expansion systems, or some combination of the two, the key is redundancy, maintainability and expandability.
Redundancy must be built into every critical mechanical system installed. An N+1 (need +1) design is typically the minimum requirements for a critical data center. If a single chiller and cooling tower are required to carry the load, an N+1 design would require you to install two. That design philosophy could extend down as far down as a user cared to carry it. There is always a tradeoff however. The higher the levels of redundancy, the higher the cost. Multiple levels of redundancy are also more complex and can lead to confusion and the potential of human errors during maintenance and testing.
Maintainability of mechanical systems can be achieved by having sufficient redundancy to shut down the systems being maintained or by having redundant components that can be shut down individually. For example, installing an additional primary chilled water pump that can serve either of two chillers is one way of achieving maintainability of an individual component, (the chilled water pumps).
Given the level of mergers and consolidations, expandability of the data center becomes a central point of the design. If additional capacity is required, mechanical systems must be set up so that they can be expanded without a shutdown. The sizing of the chilled water piping, the addition of extra valves, setting aside space in the central plant for additional chillers, cooling towers and pumps are among the issues that should be addressed in the design phase.
Designing a data center so that it can be expanded electrically requires careful planning at the beginning of the project. Extra breakers to isolate the areas where switchgear may be added in the future, setting aside space for future UPS modules, batteries, or generators, reviewing the sizing of the distribution panels, the sizing of the feeders and the addition of empty conduits for future feeders are all part of the initial electrical design process.
The type of high-speed fiber optics communications required in a data center vary greatly. Depending on the number of users and the bandwidth required, a data center could have everything from single T-1's to OC-48 Dual Sonet Rings. What doesn't vary is the need to provide redundant entrance facilities, with diverse paths and the capability to handle multiple carriers. Most data center built today want diversity in their communication carriers. Remember, a data center is totally dependent upon communications to provide access for its customers. The infamous backhoe cutting a single fiber optic line should not shut down a properly designed data center.
Designing a data center that can adapt to changing technology and user requirements is extremely challenging. It is critical that the design criteria established at the beginning of the project include the flexibility to respond to change. Redundancy, maintainability, expandability, the elimination of single point failures, all of these issues are critical in the design of a data center for the 21st century. The role of the data center is changing. The challenge is to design a data center that can respond to those changes.