Data center disasters and how to prevent them

* Physical factors in data center design

Data center managers’ most amusing anecdotes involve tragic failures of the “physical layer.” Take, for instance, the data center that was built on the second floor - to avoid flooding from broken pipes above - only to have the floor collapse in slow motion from the equipment’s weight. Data center facilities are very expensive to build and, therefore, should be built to last. But how do you predict the kind of dramatic increases in the density of computer racks seen in blade server environments? How do you manage rapid consolidation, breakneck growth and business continuity requirements?

Responsibility for the design and management of the data center has traditionally resided in the corporate facilities management group. But as data center requirements become more specialized, the task of designing and managing the facility should lie with IT. Designing a data center facility means tackling a host of complex problems:

* HVAC - Heat dissipation in extremely dense racks is dependent on many parameters, such as the type of flow (rising, front-to-back, etc.), the distance between racks and the density of each rack.

* Power - Supplying enough power for today’s ultra-dense blade servers is also difficult. One IT manager discovered that while the racks could fit 64 blades, there was only enough power for three!

* Structural integrity - Dense racks can be extremely heavy with the weight concentrated in a very narrow footprint.

* Access control - Various regulations and increased concern about security mean that the data center has to have strict personnel access policies and controls, using biometrics, IDs, tokens and so forth.

Getting all the above parameters right is very difficult, even if you start from scratch. But if you get them wrong the consequences can be serious: One IT manager told us that the heat and density in his data center had increased the rate of hardware failures (especially hard disk failures), causing extensive and frequent application downtime. Worst of all, this IT manager had no control over the facility - it was managed by the facilities management group.

This is exactly the kind of situation that arises from a facilities design that does not take into consideration the future needs of the data center. From a best-practices perspective, we believe that data center facilities should be managed and designed by IT. This does not mean changing the organizational chart so that IT is responsible for the parking lot and cafeteria - it means that facilities management does not include the data center. You might not be able to eliminate all data center disasters this way, but at least you will have some of the control - not just the blame.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Must read: 10 new UI features coming to Windows 10