While Uptime Institute’s four-tier system of rating data center reliability still provides a useful benchmarking methodology, industry professionals (including Uptime) agree that changes that have taken place since the system was first introduced demand a deeper refinement.

In a recent video posted on Digital Realty Trust’s Web site, the company’s VP of Technical Services Michael Manos critiqued the way Uptime’s system has become a “crutch” companies rely on for making decisions about the level of redundancy they incorporate in their data center designs.

“My problem here is that this industry has built up an entire set of cottage industries who still push (that) the higher the tier, the better,” Manos said in an interview. “Let’s face it: a general contractor or engineering firm is going to make a lot more money building or designing the more complex solution. I think the modern data center challenge (includes) looking at the space holistically with the operations and the application.”

The neglected factor
Many companies are too focused on initial facility design and construction expenses, forgetting to include ongoing operational practices and their cost into the reliability/investment equation. A vigorous maintenance regimen and strong procedural standards can make a lower-tier facility as reliable as a facility that’s built to a higher-tier design standard, Manos said.

Tarif Abboushi, VP of data center infrastructure at NTT America agreed.

“Far too many people in the industry only look at the infrastructure topology when they look at the tier rating and they completely forget (about) operational sustainability,” he said. Besides developing diligent operational practices, factors like an organization’s ability to man the facility around the clock with engineering staff and the facility’s geographical location affect its uptime numbers.

Concurrent maintainability
Abboushi also agreed with Manos that concurrent maintainability was a crucial design characteristic, regardless of a facility’s tier level.

“I don’t care if you call (a design) Tier 5 or Tier 3 or Tier 7,” Abboushi said. “The logic of concurrent maintainability is what matters to me. Because it tells me that I don’t need to take down any of my IT load at any time in order to perform any maintenance of the site.”

Uptime Insitute’s requirements for a Tier 3 facility include concurrent maintainability but it is only one element of the definition.

“The formal definition of Tier 3 has concurrent maintainability,” Manos said. “Tier 3 also has other design aspects. As long as you have concurrent maintainability, that’s the most important aspect of the design and you could potentially get away with lesser technology and redundancy choices and still operate it at higher levels of reliability.”

Clients wising up
Gerard Murray, engineering design principal at Kling Stubbins, has observed diligent customers that do look beyond the tiers.

“We have clients that are very interested in looking at how they can integrate different facilities and leverage that so they don’t build Tier 4,” he said. “There’s a lot more intelligence available; there’s a lot more resiliency in all the IT infrastructure and if it’s integrated with the facility … you can achieve higher reliability.”

Bruce Myatt, Murray’s colleague at Kling Stubbins, said that most data center owners he knew considered Uptime’s standards an excellent measure of reliability.

“However, many of them also like to cut and paste the UTI tier standards into a hybrid design that better suits their budgetary requirements,” he added. “The balance between costs and uptime is the basis for the now very common Tier 2+ and 3+ designs that better meet the needs of specific business models.”

Macro-redundancy
Murray pointed out another aspect that affects uptime (if only for companies with deeper pockets): regional redundancy. The bigger the number of data centers a company has, the less redundancy each data center has to have within itself.

While it may seem counterintuitive to save on redundancy within one facility by building an entirely separate one, the financial incentive in this approach lies in the rate of ROI.

“A lot of people are interested in that because they can drive up their utilization,” Murray said. “If you have totally redundant systems and it’s less than half-loaded, the equipment isn’t operated at its optimum efficiency.”

Investment in UPS units, for example, does not start paying off until there is a main-feed interruption, Abboushi pointed out. “The other factor that comes into play is that UPS’s generally (are inefficient) and the efficiency increases as the load increases.”

A post-convergence tier system?
Generally, the tier classification system provides a useful framework for a discussion of data center reliability, Abboushi said. “These four benchmarks, that the Institute has articulated, I think are very good benchmarks. I don’t think (Manos) was being critical of the Institute. I do think he was being critical of the way the Institute’s standards are massaged by people in the industry.”

Manos said that he has used the Institute’s system in his work and will continue to use it. The system, however, only addresses “one aspect of the evolving data center space” and has to be treated as such. “It has become a victim of its own success, whereby the facilities requirements or Tier level is what has been driven into the minds of executives.

“The data center, today more than ever before is a whirling blend of IT, facilities, and operations. Like it or not, the evolution of the data center has poured them into the same bowl and has started mixing and blending. I personally believe that we have just begun the modern baking process. It will be interesting to see what cooks up in the end.”

Related news: Digital Realty Trust splashes $44.5 million on two Californian data centers
Related whitepaper: A review of data center tier classifications
Related feature: Mike Manos Microsoft’s (former) data center shipping magnate