With the prevalent and continued growth of technology, locating and building a server room in an existing facility...
is very common. What issues need to be addressed when undergoing the server room design process? This two-part article will help identify the main considerations by defining individual requirements and looking at various physical building systems.
What is a server room? There are a number of definitions, but they are too numerous to cover here. I will focus on spaces dedicated solely to housing technology equipment and its supporting infrastructure inside an existing building. The building is expected to have other uses besides the technology equipment, whether that is office space, storage space, a health-care facility, institutional space or any number of other types of uses.
It is critical to understand there is no one-size-fits-all solution to a server room design. There is a growing trend toward standardized modular designs or more of a manufactured product to create the server room environment, as opposed to the traditional unique, "constructed" environment. There are many situations when a standard manufactured product is the right design for a server room, just as there are times when it is not.
Due to past failures and problems in server rooms, there are many best practices that should be included in just about every design. But these are parts of overall design solution, and do not override the importance of designing the right solution for the unique needs of a business.
Although there can be a long list of factors to consider when designing a server room, we will break it into two general categories: Defining the requirements and building system considerations.
Defining the requirements entails load density, resiliency, and growth, modularity and flexibility.
Find out your load density
Traditionally, load density is expressed in watts per square foot (W/sf). For server rooms, although W/sf plays a role, the use of both the maximum and average kilowatt per rack (kW/rack) is more meaningful. These two numbers drive much of the electrical and mechanical system requirements. For example, the electrical and mechanical systems to support a space with an average of 3.5 kW per rack and a maximum of 5 kW/rack are very different from those to support an average of 10 kW per rack with a maximum of 12 kW/rack.
This density drives not only the system capacities but also the configurations. Examples where the density affects the mechanical system configuration are decisions such as close-coupled or centralized cooling, overhead or under floor cooling, segregated hot and cold air streams or open airflows. The impact on energy efficiency also drives these system configurations. For example, take the hot and cold air segregation. At higher kW/rack densities, this is necessary for proper operation. At lower kW/rack densities, the end-use equipment can be cooled without the air segregation, although there is a significant gain in energy efficiency by segregating the hot and cold air. But is segregating the airflows required for a server room design? That depends on the unique business needs, weighing the higher installation cost with the payback potential.
There is a common discrepancy seen in server rooms where building system capacities are mismatched. Often this is the mechanical system limiting the load. The load density must be applied as a universal number holistically across every component and system. Otherwise there are wasted costs and stranded capacities.
Building server rooms for resiliency
The next important step is to determine the resiliency or the level or redundancy required. This will lead to the total required installed capacity, or -- translating it to dollars -- how much capacity must be purchased. The resiliency is more important than determining the total required capacity; the resiliency requirement gives the criteria to determine total capacity, which the appropriate design team should calculate.
Two factors drive the resiliency requirement: the business requirements and the opportunity to maintain the critical supporting equipment. The business requirements discussion should involve the end users and can be challenging. Some typical topics to resolve are:
- Do these systems need to be running at all times?
- What is the impact to any business activities if the servers are offline?
- Are these systems customer-facing?
- Is there redundancy from a software perspective where another site can pick up the server load, thus allowing this site to have outage?
The costs and the complexity of building systems rise rapidly as they become more redundant and more complex, so an honest evaluation of whether a system can be offline for planned maintenance is critical.
All mechanical, electrical and plumbing equipment needs periodic scheduled maintenance -- preventative or predictive, unplanned or restorative. Scheduling maintenance around the end user's needs reduces the required level of redundancy and cuts costs. Without the discussions with end users, the resiliency of the systems is often either underdesigned or overdesigned for the business needs.
Similar to the holistic approach to the density, the right level of redundancy must be applied to every piece of equipment and system supporting the critical load. This helps avoid a common pitfall where one system or component limits the maintenance opportunities or resiliency of the entire supporting infrastructure. For example, it is not uncommon to have to defer critical maintenance on an automatic transfer switch with no bypass because performing this maintenance requires a complete shutdown, causing an unacceptable outage to the critical load. Another common situation is to defer maintenance because it is assumed the critical load cannot be disrupted, when in fact it could handle a scheduled disruption.
There is no right answer for what resiliency a system should meet; the answer is to design the server room right from the start to meet the business needs.
Tackling growth, modularity and flexibility
The final piece of the puzzle defining the requirements is deciding what will be the initial load and how the load will grow over time. Some of the key questions to ask are:
- Is this a permanent installation or something temporary to bridge a few years until a more permanent solution is reached?
- What is the typical technology refresh cycle?
- Does there need to be "swing space" to handle incoming technology while transitioning from one system to another?
- Can the building systems -- including the physical space -- be broken into smaller segments to allow it to be built in modular blocks?
A modular approach has grown in popularity in recent years and can help pace the growth and ease the technology refresh. But it comes at a cost so, like everything discussed above, it is not always the right solution. There are many more ways to support flexibility and modularity than can be discussed here. The business needs and the requirements for the space will determine the right solution.
These items establish the baseline for designing a server room. In the part two, we will look at the key factors to consider in regards to the specific building systems -- the physical space, mechanical systems, electrical systems, ancillary systems -- when designing a server room.
About the author: Tom Langran is a licensed professional engineer with more than 15 years of experience in building design and construction. Tom has a master's degree in electrical engineering and construction management. Tom has designed and managed data center projects for financial institutions and telecommunications carriers. He lives and works in Seattle.