Article: Debunking Data Centers
By Tom Deaderick, DatacenterJournal.com
April 21, 2009
Clean white server rooms, security cameras, bio-metric scanners, enormous cooling units, uninterruptible power supplies and massive generators make professional data centers look more impressive than your server room, but are they?
Touring a professional data center can be a reassuring experience, but what if the view behind the curtains isn’t as impressive? There are many reasons organizations outsource data centers. Ramping up an internal server room is expensive and achieving true A/B power separation with complete redundancy in a building that’s not specifically designed to be a data center is easier said than done. An in-sourced data center can also limit a growing organization’s choices, with the investment tying the organization to a building or location that no longer fits the business plan.
The best location for your enterprise infrastructure is determined by one factor above all others – reliability. Spend a lot on infrastructure, with the most powerful processors, SAN drives and management applications, or spend just enough. When the power’s off neither creates value and the most expensive infrastructure will just get you fired faster when it isn’t working.
Reliability is the most important factor. If the system you designed is operational when the business needs it, it’s priceless, if it’s not, worthless.
A variety of factors impact reliability, but the best place to start is the Uptime Institute. The Uptime Institute developed four categories to evaluate the reliability of professional data centers as well as the server room in your office. The categories are “tiers” and four simple descriptions allow benchmarking of any data center’s expected reliability. Tier 1 is the lowest, Tier IV the highest. The costs required to achieve each successive Tier level are more exponential than linear. The Tier standards are simple but exacting.
Google “tier III data center” or “tier IV data center” and many commercial data centers are displayed. Unfortunately, the claims of nearly all these commercial data centers are exaggerated. Julian Kudritzki, Certification Manager for Uptime Institute Professional Services, relates that two data centers, which had previously claimed Tier IV or ‘close to it’, completed the Certification process last December. Both received a Tier I Certification. Understandably, neither organization chose to publish the true Tier level of their data center.
Consider the clients of these exaggerated-reliability data centers. Imagine the reactions of those IT managers who made the decision to outsource data center operations with the intention of providing improved reliability above their own internal server room when they discover their commercial, professional, data center actually has a number of single-points-of-failure just like their own server room.
Hold data center providers to a higher standard. In fact, just hold them to the standard they already claim. If a data center operator is trying to convince you to move your enterprise infrastructure into their facility ask to see their Certification, or save time and check out the interactive world map of commercial Tier-Certified data centers at http://professionalservices.uptimeinstitute.com/TierServiceProvidersMap.htm.
There are many other factors worth considering in the decision between in-sourcing and outsourcing the data center.
Location, location, location
Every IT Manager knows the most likely cause of a service disruption is not a tornado, hurricane, fire or a flood. Most outages are the result of human error or actions. Humans do dangerous unpredictable things as a few minutes on YouTube clearly illustrates. Humans in the vicinity of system infrastructure create more risk. This is usually a challenge for the in-sourced data center/server room because the majority of the surrounding building supports people from other departments like Human Resources, etc. There are people everywhere, surrounding the in-sourced server room. Any one of them could initiate a disruptive incident at any time. How many candles, floor heaters and consumer electronics are in the building with your data center? What happens in the building over the weekend?
So if humans are dangerous and unpredictable, are out-sourced data centers better? Not necessarily. It is surprising how many commercial data centers are located in buildings containing hundreds of people. People you know nothing about. Most often, these data centers will be located in the basement. Sir Isaac Newton, famous for the discovery of gravity also had some guidance for those considering locating a data center in the basement of a building with bathrooms on each floor above, “I don’t recommend it”, I think he said. Good advice.
A data center should be a data center only. There should not be a call center, or office operations or anything else in the building. There’s an irony to paying for HVAC and UPS equipment to create consistency in a tiny environment surrounded by the most unpredictable creatures on the planet.
Expanded perspective on location
Extend this theory and find another important consideration. Should a data center be located in metropolitan areas, bustling with people? Most data centers are in large metropolitan areas. The risks of thousands of people initiating a disruptive event (chemical spill, fires, floods and even terrorist attacks) are obviously greater than the risk of a smaller number of people. If you’re considering outsourcing the data center, keep this in mind. You could be moving from the frying pan into the fire.
We consulted a food services organization on improvements to their data center. They weren’t interested in out-sourcing their data center, so we provided recommendations to improve the infrastructure to practical limits. Every improvement recommended would not remedy one glaring weakness however. The data center was located in an active shipping warehouse with a rail line less than a hundred yards away. The location was convenient for shipping product, but a potential catastrophe for the data center. Every data center manager should review the statistics for train derailments on this wiki (http://en.wikipedia.org/wiki/Derailment). Check with your region’s rail carriers. They will tell you they frequently carry very hazardous materials.
On June 28, 2004, a Burlington Northern Train with 123 empty cars and one white tanker containing 15,000 gallons of liquid chlorine crashed into a Union Pacific train. The chlorine trapped dozens of people for as long as eight hours in homes where the chlorine gas was so concentrated that it dissolved metal car keys. The horrifying story is the subject of an episode of the series “I Survived” on the Biography Channel, and a complete narrative is available online in this reader’s digest article (http://www.rd.com/your-america-inspiring-people-and-stories/trapped-in-a-toxic-cloud/article27195.html).
That food services data center is just one bad day away from disaster.
One bad day.
The central insight here is simply this, “Question your assumptions”.
Your internal server room/data center has weaknesses, single-points-of-failure that will eventually bring your systems down. Admit it and get professional advice on costs to correct the highest risk or most frequently occurring failures.
With these estimates in hand seek pricing from professional data centers. Visit the Uptime Institute’s web site and identify a data center with credentials that are backed up by someone other than the data center operator.
Review your data center objectively against the professional facilities on your short list.
Which choice is truly the safest and most reliable location?
Remember, there are only two states for business infrastructure. In the “off” state infrastructure is worthless.
About the Author
Tom Deaderick, Director of Business Development, OnePartner. Deaderick has nearly 20 years of information technology experience, with Fortune 500 companies like Eastman Chemical and Raytheon, to start-up firms such as e-STEEL, Blackfin and Intellithought, where he most recently served as the company’s President. At OnePartner, Tom oversaw the completion of the first Certified Tier III disaster recovery and commercial hosting facility in America, OnePartner’s Advanced Technology and Applications Center (ATAC). In his current position he manages day-to-day operations of the ATAC data center which includes coordinating with clients and business partners on data center storage needs, overseeing the data center’s maintenance, and more. Current ATAC clients include large financial and healthcare organizations, such as: Bank of Tennessee, CareSpark and Holston Medical Group.