Skip to Content

A Smarter Algorithm Could Cut Energy Use in Data Centers by 35 Percent

Storing video and other files more intelligently reduces the demand on servers in a data center.
April 16, 2013

New research suggests that data centers could significantly cut their electricity usage simply by storing fewer copies of files, especially videos.

For now the work is theoretical, but over the next year, researchers at Alcatel-Lucent’s Bell Labs and MIT plan to test the idea, with an eye to eventually commercializing the technology. It could be implemented as software within existing facilities. “This approach is a very promising way to improve the efficiency of data centers,” says Emina Soljanin, a researcher at Bell Labs who participated in the work. “It is not a panacea, but it is significant, and there is no particular reason that it couldn’t be commercialized fairly quickly.”

With the new technology, any individual data center could be expected to save 35 percent in capacity and electricity costs—about $2.8 million a year or $18 million over the lifetime of the center, says Muriel Médard, a professor at MIT’s Research Laboratory of Electronics, who led the work and recently conducted the cost analysis.

So-called storage area networks within data center servers rely on a tremendous amount of redundancy to make sure that downloading videos and other content is a smooth, unbroken experience for consumers. Portions of a given video are stored on different disk drives in a data center, with each sequential piece cued up and buffered on your computer shortly before it’s needed. In addition, copies of each portion are stored on different drives, to provide a backup in case any single drive is jammed up. A single data center often serves millions of video requests at the same time.

The new technology, called network coding, cuts way back on the redundancy without sacrificing the smooth experience. Algorithms transform the data that makes up a video into a series of mathematical functions that can, if needed, be solved not just for that piece of the video, but also for different parts. This provides a form of backup that doesn’t rely on keeping complete copies of the data. Software at the data center could simply encode the data as it is stored and decode it as consumers request it.

Médard’s group previously proposed a similar technique for boosting wireless bandwidth (see “A Bandwidth Breakthrough”). That technology deals with a different problem: wireless networks waste a lot of bandwidth on back-and-forth traffic to recover dropped portions of a signal, called packets. If mathematical functions describing those packets are sent in place of the packets themselves, it becomes unnecessary to re-send a dropped packet; a mobile device can solve for the missing packet with minimal processing. That technology, which improves capacity up to tenfold, is currently being licensed to wireless carriers, she says.

Between the electricity needed to power computers and the air conditioning required to cool them, data centers worldwide consume so much energy that by 2020 they will cause more greenhouse-gas emissions than global air travel, according to the consulting firm McKinsey.

Smarter software to manage them has already proved to be a huge boon (see “A New Net”). Many companies are building data centers that use renewable energy and smarter energy management systems (see “The Little Secrets Behind Apple’s Green Data Centers”). And there are a number of ways to make chips and software operate more efficiently (see “Rethinking Energy Use in Data Centers”). But network coding could make a big contribution by cutting down on the extra disk drives—each needing energy and cooling—that cloud storage providers now rely on to ensure reliability.

This is not the first time that network coding has been proposed for data centers. But past work was geared toward recovering lost data. In this case, Médard says, “we have considered the use of coding to improve performance under normal operating conditions, with enhanced reliability a natural by-product.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.