A recent article in The New York Times, titled “Power, pollution and the Internet”, reminded me of similar events in the past. Such an article could easily have been published in the early 1900s, when assembly lines introduced mass production to the world and transformed the automotive industry, for example.

It doesn’t take a leap to imagine outrage at these evil self-propelled personal transportation vehicles we call automobiles that could some day “take over the roads”, replacing the railway and the beloved horse and carriage. Wasteful people will want to use this gas-consuming, noisy and dangerous transportation device to travel with greater speed and to more destinations. “It’s not sustainable,” the articles would claim. “They’re going to hit a brick wall.”

Today, we refer to that era as “the second industrial revolution”, while referring to the internet age, with millions of data centers worldwide, as the “digital revolution”. The switch from steam-driven trains to combustion-engine automobiles represented an instant 500% improvement in efficiency, and personal miles per gallon (pmpg) has impcorved almost 300% in the last 100 years.

The data center industry has gained efficiency improvements far beyond this benchmark just over the past 10 years. At the same time, the Internet and the infrastructure that supports it have dramatically simplified nearly every aspect of our lives: from banking to shopping to the way we communicate.

As much as I disagree with most points The Times article makes (and the dated information that supports them), several industry experts the author quotes zero in on an important issue. Yes, we generate several quintillion bytes of data each day, and yes, servers still consume electricity despite technological improvements. And yes, we do feel that we have a God-given right to access digital content at any time and without interruptions. In this context, data centers are still being designed inefficiently, and this is particularly true for the physical facilities infrastructure: power and cooling. The industry’s fundamental challenge is to break this paradigm. We need to rethink legacy designs and, more specifically, electrical-infrastructure designs that have been around for decades – the highly inefficient ones that don’t meet performance needs of today’s mission-critical facilities.

An ancient tradition

Much of the industry still treats data center design and construction as a craft passed down through generations of engineers, each adding her or his own design preferences. As a result, no two data centers look alike and no two data centers can be operated or maintained the same way. The legacy lives on because data centers are managed by extremely risk-averse individuals, whose careers are on the line every time they introduce something new. The repercussions of failure far outweigh the rewards of innovative success.

Too often do we see a request for proposal  (RFP)or a quote, with a boiler-plate data center specification based on a 10- , 15- , or even a 20-year-old design. Once on paper, the specifications are rarely questioned. They pass through many layers however: from the customer’s customer, to the customer, to the architect, to the engineer, to the contractor and, finally, to the vendor. By then, questioning the requirement or proposing an alternative solution is too overwhelming for anyone involved.

Unnecessary over-provisioning

Let’s take for example the requirement that an uninterruptible power supply (UPS) has to provide 15 minutes of ride-through time. This requirement originated from a 1952 IBM Series 701 Mainframe specification. It said the mainframe needed 15 minutes to conduct a graceful shutdown. Backup generators were not generally in use at the time. To this day, most data center RFPs, however, require 15 minutes of ride-through, even though backup generators are almost always deployed in medium-size and large data centers, and their five-to-eight-second startup time makes ride-through times longer than 30 seconds unnecessary.

There are many examples like this, and they all directly contribute to the wasteful practice we in the data center industry refer to as “over-provisioning“.

Thousands of inventions have brought to the market technologies that promise massive efficiency improvements over the past 10-15 years. Yet, as a whole, the industry’s efficiency has improved only incrementally, because of this large group of risk-averse individuals who refuse to question business-as-usual.

Some of the data center industry’s most forward-thinking innovators have completely changed conventional deployment paradigms by introducing standardized modular designs and supporting them with contemporary lean manufacturing systems. Many of them are new players in a crowded market, but they enable data center operators to get services to market faster and, as a result, are enjoying some of the highest growth rates in the business. Data center design and deployment should not be a honed craft passed down from generation to generation, but rather a lean and fast service offering, where improvement is easy to manage, implement and repeat.

Fresh, forward-thinking approaches to data center design will make critical facilities less wasteful, leaving no fuel for criticism, such as the ideas The Times article put forth. Luckily, a new breed of suppliers is infiltrating the ranks of yesterday’s solutions and reaping the rewards of being able to get the industry’s innovators to market faster and more efficiently.

This article first appeared in DatacenterDynamics FOCUS magazine. Click here for free subscription.