Let's tell the whole story on data centers

Let's tell the whole story on data centers

Editor's Note: To learn about real-world applications of big data and hear from Kathrin Winkler, be sure to check out [email protected], November 12-13.

Did you see the New York Times article on Sunday, September 23 titled "Power, Pollution, and the Internet"? If so, are you one of the three people of my acquaintance who did not send it to me.

Even my mother asked. This is what I told her: It's really not new, at least to those of us in the industry. And it's only half the story. I'm just sayin'...

Yes, demand for information technology is growing by leaps and bounds, and yes, it takes power to drive it. It's also true that availability ranks far higher than energy efficiency in any data center manager's priority list. As one IT professional put it to me on Friday, "At the end of the day, I'm paid to keep our infrastructure running."

And yes, some of the performance demand is to provide sub-second response time to view YouTube videos. But some of it is to keep planes from running into each other, to complete genome analyses, to manage traffic, to diagnose illnesses, to predict pandemics, and to warn of oncoming tsunamis. This is one of the places where the article falls short — in its zeal to highlight the waste, it undercuts its own story by neglecting the amazing value provided by information technology. That value may be economic — try comparing the value to the economy per kWh of IT versus other energy uses — or it may be human, from health care to education to personal support (and, yes, entertainment) — or it may be environmental, preventing over-fertilization of fields, or siting wind mills or solar panels.

Don't get me wrong — I'm not saying there isn't waste. In fact, I once coined the term "d-waste," or "data waste" to represent all the crap we've saved that we don't need, and all the excess copies of the data that we do need. There are indeed frivolous uses of data; the article is right about that. But it's not the whole story. And the flow of data isn't just from consumers. It's from cars, traffic lights, cameras, MRI machines, weather satellites, space explorers, and so much more. I'm just sayin'…

I also wish the article had given its due to the things that the industry is doing to be more efficient. I've said this before: By my calculation, if the automobile industry had gotten efficient as quickly as the IT industry in the last 40 years, we'd be getting 450,000 miles to the gallon. Admittedly, it hasn't been enough to offset the growth in demand. But look what we're getting for it!

And the industry is not sitting still. First of all, companies are collaborating in organizations such as The Green Grid, which is dedicated to resource efficient IT. And it's not just the server vendors — it's the storage and networking vendors, data center operators, end users, facility architects, academia and more. We've developed metrics such as PUE (Power Usage Effectiveness), CUE (Carbon Usage Effectiveness), and WUE (Water Usage Effectiveness) to help find — and then eradicate — waste in the operation of data centers. We've created educational tools to help data center operators — who after all, only benefit from reducing cost.

Then there are the technologies and practices that are being deployed both to manage the growth in demand and to serve that demand more efficiently. Yes, most disks spin — but there is increasing deployment of solid-state drives, whose efficiency and cost are rapidly descending the learning curve. There are software technologies such as EMC's FAST (Fully Automated Storage Tiering) that put only the most critical data on the faster (and thus higher energy) disks. Or data de-duplication, which gets rid of so much of the "d-waste".

And yes, the article mentioned virtualization, but barely, and didn't even touch on the scale of the impact it can have on utilization — as we know from the 4x improvement in our very own data centers. In fact, the discussion of Cloud Computing, while rightly pointing out that cloud does eventually use actual hardware, really neglected the subject of how the use of private and hybrid clouds can make enterprise infrastructures more efficient by reducing over-provisioning; it's not just about users storing old email, as the article suggested.

Do we have work to do? Yes. We need to get even more efficient both in the technologies themselves and in how we use them. We need to find other sources of reliable backup power so we can get away from those diesel generators. We need to be able to show users what the impact is of their actions.

But we've come a long way, we're having a positive impact, and we're working to improve. I'm just sayin'...

Data center image by Oleksiy Mark via Shutterstock