Back to blog

Design Approach for Data Center Engineering


I recently attended a workshop and certification program sponsored by the Department of Energy for Data Center Energy Analysis.  Though the content of the program was more geared to generalist training, there were a couple things that I reaffirmed in Morrissey Engineering’s design approach for data center engineering as well as our approach for other high performance/high reliability building types.

Less is often better – Design of high reliability facilities has always been about balancing redundancy with simplicity.  A meaningful design strives to create required redundancy with the fewest right sized components possible.  If a client's uptime needs dictate an N+1 solution, N+4 is not better.  In fact, it can be much worse.  Oversized equipment usually means equipment operates in its most inefficient manner.  This is a big penalty for our building Owners.  Not only did they pay a cost for the extra equipment, they have ongoing maintenance costs and significant distribution system losses that are a seldom known portion of the utility bills.

Realizing that for most facility and IT managers, energy use and energy costs are not big on their radar screen today, one can argue that they should be or soon will be.  Many of our clients enjoy power costs around $.055 per KWH.  This means that for every 1000 watts of wasted energy, the annual utility cost is $500.  Even with “cheap” power, losses due to poor load density in a facility can easily be tens of thousands of dollars in utility costs and wasted resources annually.

As much as some may not approve of government involvement in the buildings we occupy, there is a strong push to raise the awareness of building efficiency through benchmarking.  The most common data center metric of efficiency is the Power Usage Effectiveness (PUE) performance metric.  PUE is a ratio of total energy used to energy used for IT equipment.  It certainly doesn’t tell the whole story about a data centers efficiency but industry experts are predicting a government mandate for publishing PUE by Major Corporation and government data centers.  This same discussion is occurring with all major buildings with the concept of a National Energy Code.  The thinking is that if a building's performance is known, the real estate market will factor that into the value of a building, thus making inefficient buildings less valuable compared to high performers.

Like most metrics, there can often be an unintended result.  This is true for PUE and other facility benchmarking tools.  A misaligned marketing tendency exists to use PUE as a way to distinguish one facility to another.  Since the design topology has such an impact on PUE, it’s impossible to accurately use it to compare one facility to another.  A simple facility designed with N components and no redundancy will logically have a higher efficiency compared to a 2N facility that has complete redundancy of components.  The true use of PUE is in first creating a baseline to get a sense of how a facility is operating, and then comparing that baseline of data to changes made to a facility as a feedback mechanism.  Feedback is common in so many circles yet we have few meaningful and accurate feedback loops in the design of buildings.

For now, there appears to be no consistency in the way PUE is measured.   It’s difficult to do correctly and in a meaningful way that provides benefits.  Further confusion was created recently with the release of PUE Version 2 which states that building energy (kWh) should be used for the ratio, not power (KW). 

Speaking of rating systems, we here at Morrissey Engineering have been working on a decision tool for data centers that provides a numeric value to a facility's attributes by using statistical analysis.  When complete, this tool will be used when working with our Clients to determine how best to update their facilities, or to determine what attributes to look for in a new facility.  It’s a large project but one that will allow us to better communicate design decisions and help our Clients get the best product to meet their needs.

Detail is a deal – As the economy continues to struggle, we see more pressure in our industry to reduce scope and design detail to lower the cost of design.  It seems a lot like preparing for an exam in school.  The more study of the problems we do, the better our responses are.  For engineering of data centers, there are so many design details that make a great end product yet many times they don’t show up on a Client wish list or Tier rating system. 

As a young student pilot, I had an instructor who once told me that when things don’t go as planned, you don’t want to say “Oh no, this unexpected thing happened; now what am I going to do.”  Instead the response should be “Oh no, this unexpected thing happened; now here’s what I’m going to do.”  I’ve used that same thought process as a guide to putting meaningful detail into the data centers I’ve worked on. 

Design documents that are clear, comprehensive, and are influenced by years of experience absolutely provide the best end product and best value to our Clients.