Building Failsafe Measures during the Design Phase of Data Centers

August 20, 2013

By Michael Guta - Contributing Writer

Data centers are the heart of the digital world we all seem to be living in, and if an unforeseen disaster takes down the multiple contingency controls that are in place to protect the center, the damage can add up to millions of dollars every minute it is down. With new datacenters costing hundreds of millions of dollars, every possible scenario is given a solution before construction of the facility even begins. Although great measures are taken during the design phase, incidents still take place which can slow or even stop operations long after the center opens.

In theory, each new data center should be the best facility housing the servers that hum day and night, but this is not the case. Many different constraints determine how the center will be built, and if budget is not an option then it is possible to build a fault-free data center. If, on the other hand, the designer is faced with budget shortages, the cutting of corners will manifest itself sometime in the future in the form of decreasing productivity or even catastrophic failure.

A data center’s primary goal is to stay operational no matter what type of event takes place, and to ensure it stays operational, multiple redundancies must be put in place. The complexity of a center housing thousands of servers generating high temperatures, with multifaceted cooling systems, multiple power sources, and single isolation points of failure in a 24/7 operation is overwhelming, yet they remain operational; thus, everything hinges on the design of the center.

Every system has a finite probability of failure, meaning it will fail at some point, and understanding how this failure will affect the data center is as important as preventing it. So the design of a data center should:

  • Simplify the process;
  • Isolate failure to single points;
  • Have several layers of redundancy, but not so much switching slows or downs operations;
  • Choose the right goal of scalability;
  • Select the correct power density; and
  • Plumbing – water is as important as power if it is used for cooling. Although very rare measures have to be put in place just in case there is an outage.

By now, we all agree the design of a data center is an extremely complex process, and depending on the size and capacity it can get even more complex. Creating a center with the allotted budget is almost always impossible, because customers generally want more than they can afford, but addressing some of the most common mistakes before construction even begins can go a long way in ensuring it stays operational while facing the most unlikely scenarios.

Edited by Rich Steeves