Part 3 in a 4-part series
Myth #1: Large Data Centers Aren’t Green Or Efficient
A common misconception about large data centers is that they’re neither green nor efficient. Compare a bank that runs five smaller internal data centers with a similarly sized bank that runs a single, large data center. From the standpoint of efficiency, power consumption and cost, the latter wins every time.
Likewise, a large data center can be made more efficient than a small one. That’s because any data center, large or small, faces certain constants: supply of electrical power, facilities operation, support personnel, security guards, and more. These operating costs remain relatively fixed, regardless of facility size.
Furthermore, smaller environments require greater equipment redundancy, especially for cooling, UPS and generators. Finally, dedicated personnel and processes must be deployed to actively maintain an optimal load for the data center, avoiding excessive power consumption or underutilized infrastructure.
From their conception, larger facilities, if designed correctly, offer the opportunityfor better operating efficiencies, producing more energy while consuming fewer resources. But providers of large, colocated data centers take the efficiency results a step further. Their size means they offer economies of scale – across global locations – that few, if any customers could cost-effectively match on their own.
Furthermore, these providers are effectively obliged to commit publicly to be stewards for responsible energy consumption. Consider that while the average data center has a power usage effectiveness (PUE) of 2.5 according to widely cited statistics from organizations such as The Green Grid and Uptime Institute, over the past two years, Equinix’s new builds have achieved an average design PUE of 1.6.
Myth #2: All Organizations Compute PUE The Same Way
The formula for computing power usage effectiveness is simple: Divide the kilowatt-hours that a facility consumes by the total kilowatt-hours that the facility’s IT equipment uses (or in the case of data center providers, what customers’ IT equipment consumes). In a perfect world, the result would be a PUE of one. But lighting, humidifiers, and cooling equipment all diminish effectiveness.
Some organizations get close to the ideal. Equinix, a carrier-neutral provider that doesn’t dictate what its customers install on its premises, maintains an average design PUE of 1.6 for facilities built in the past two years. That is testimony to Equinix’s heavy investments in facilities that benefit from the latest cooling capabilities.
Of course, the current PUE metric itself can be misunderstood when attemptingto use it to make comparisons between sites. Different geographical locations, climate, occupancy rates and design can cause variances. Factor in the time of year, weather and the nature of the applications running in the data center itself (the IT load) and the current static ‘snapshot’ PUE measurement loses some of its value as a comparison tool.
To help clarify such discussions, organizations such as The Green Grid, of which Equinix is a member, are proposing, through a published whitepaper, refinements to the current PUE method in order to address some of the current limitations. These proposals include defining a set of new PUEcategories that reference a yearly average performance rather than a snapshot’ measurement.
With luck,the industry will soon have a much better way on a consistent way to compute, and thus compare, efficiency.