Google

How Google and Yahoo improved data center energy efficiency

Running large data centers is expensive and getting worse. No one knows that more than Google and Yahoo. At a panel on green data centers at MIT’s Emerging Technologies Conference this week, Yahoo’s senior director of data center engineering and operations, Scott Noteboom, described how Yahoo has grown its server plant by 12 times since 2005. Google’s head of data center research and development, Chris Malone, explained that Google has had to come up with its own server designs to accommodate the ultra-high densities of their data centers. Both have a maniacal focus on improving data center energy efficiency.

Yet both firms are doing things that ordinary enterprises can learn from:

1. Reconsider your business continuity plan. Yahoo is eliminating UPSes and backup generators. Instead, they are architecting each data center as a backup to the others. That eliminates UPSes as a significant power loss. They’ve had to install software that restarts servers in a controlled fashion after power resumes, however. Google took a different tack: eliminating larger UPSes in favor of small battery backups on each server motherboard. Either way, both data center approaches yield significantly greater energy efficiency.

2. Consider the weather when you site a data center. Yahoo is no longer building raised-floor data centers. By locating in more moderate climes, they’re able to utilize prebuilt warehouse-type buildings that use extensive ground-level air intake of cooler ambient air and roof-level exhaust of hot air.

3. Raise the temperature and humidity. Modern servers can run warmer and moister than you might think. A lot of conventional wisdom about server environments stems from the mainframe days, when you didn’t want punch cards wilting.

4. The big win in facilities power reduction in the data center is in cooling. Malone points out that typical chiller systems are only 60% efficient; Google has gotten to 90% by switching to evaporative cooling. Focusing on electrical transmission issues is good but yields much less improvement.

Google and Yahoo make extensive use of virtualization to improve capacity utilization and reduce the overall load. Noteboom and Malone stressed that senior IT managers need to make these issues part of the discussion with system architects, project managers and developers. All too often, those folks start from the premise of buying a certain kind of server for a certain kind of app, rather than asking how existing capacity could be deployed to meet that need.

And the result of all those efforts at Google and Yahoo? Dramatic improvements in power efficiency. The typical metric for this is power utilization efficiency, which means the ratio of total power consumed by the data center to total power consumed by just the servers themselves. Most contemporary data centers are in the 2.0-2.5 range – for every watt of power that a server uses to compute, they burn more than a watt in transmission loss, battery loss and cooling. Google is at 1.19 and Yahoo is targeting 1.03 in its next-gen data centers. In other words, if you’re typical, you’re spending twice what they are on power per server.

Is Google Chrome OS a turning point or yawner?

Nothing fuels the tech press like wars. So I have no doubt that we will see an endless string of reports from the battlefield over the next many months on the struggle between Google and Microsoft for world dominance. Wake me up when it’s all over.

Google’s announcement of its intention to field a Linux based, browser-centric operating system for netbooks and eventually most client machines comes at a strange time in the history of operating systems. Now, operating systems matter less than ever.

Once upon a time, your choice of operating system dictated your choice of processor, which often meant your choice of hardware vendor. It also dictated what applications were available to you and from whom. The only mainstream operating system that you can still say that about is the Mac OS.

Chrome OS will do none of those things. It will run on all the same hardware that Windows runs on, plus ARM-based systems. It will run browser-based applications, which will also run on any other browser. Undoubtedly, Google Apps will work really well on Chrome OS. Google is also suggesting it will deliver better security, easier configuration and quicker performance than the incumbents.

History (especially of desktop Linux) suggests that it will have to be a quantum leap better at all those things to make a dent. We can only hope they achieve that, since regardless of who delivers those attributes, they are desirable. But more to the point, Chrome OS comes at a moment when desktop operating systems are themselves in danger of becoming a big so what.

By the time Chrome OS shows up, desktop chips will likely have hardware-assisted virtualization. Increasingly, hypervisors will determine the performance characteristics of the system – how well they manage memory, how well they interact with CPUs, etc. With storage and applications increasingly being network-driven, the operating system’s chief function may well become the user interface. How security functions will be parceled out between hypervisor and operating system is perhaps an open question, so that function may still be crucial.

Nonetheless, the world Chrome OS makes its debut in will be one where operating systems can be swapped willy-nilly, where applications don’t care what OS they run on, and where, frankly, users may not either.

Google’s main impact may well be nontechnical, forcing Microsoft to drop prices and (just) maybe to improve Windows. Call me a contrarian or curmudgeon, but I think this development is more important to shareholders of both combatants than IT managers. How about it, CIOs? Game changer or no biggie?

Academics at MIT CIO Symposium advise on innovation, future of IT

CIOs look to the MIT CIO Symposium for information on management, technology and innovation. Those in attendance at the academic panel held in Kresge Auditorium in Cambridge, Mass., enjoyed a healthy helping of all three, as distinguished researchers from MIT’s business centers offered ample insight into how successful organizations leverage technology to increase innovation and profit. Over the course of the hour, the audience heard about the power of collective thinking, the impact of Web 2.0 tools behind the firewall and the methodologies for innovation that have served to differentiate IT giants like Google from their competitors.

[From left to right: Prof. Erik Brynjolfsson, Gary Beach, Prof. Thomas Malone, Dr. Jeanne Ross]

The moderator for the panel, publisher emeritus of CIO magazine Gary Beach, didn’t waste any time, asking each academic what “the next big thing” in IT was. Professor Thomas Malone, Director of MIT’s Center for Collective Intelligence, noted immediately that the “elephant on the table is cloud computing.” In his opinion, “It may well be the next being thing in the hardware progression.” He chose, however, to focus on the power of collective intelligence.

His choice may not be surprising, given his research, but his coinage of the term crowd computing to describe distributed online collective intelligence turned to solving problems drew appreciative chuckles from the crowd. In Malone’s view, the answers of the many, or so-called “wisdom of the crowds,” is a powerful tool for organizations seeking answers to tough questions. Malone noted Twitter and Innocentive as two examples of the concept.

Dr. Jeanne Ross from MIT’s Center for Information Systems Research (CISR) chose to focus on the digitization of organizational resources, stating that in her view, only “about 2% of global companies have nailed the concept of a digitized platform.” She said here are two things IT does well: standardizing and integrating business processes. Organizations of all types will gain the most from their IT investments by focusing in these areas.

Professor Erik Brynjolfsson, author of the forthcoming Wired for Innovation: How Information Technology is Reshaping the Economy, sees opportunity in the downturn. As he noted, “the lion’s share of Fortune 500 companies were founded in earlier economic disruptions.” Brynjolfsson calls today’s recession the “great restructuring.” In that trend, he sees three key elements: Experimentation, measurement and scale

Each of these elements is substantially enabled by innovations in information technology, like A/B testing, Web and data analytics and cloud computing or enterprise resource planning systems. Brynjolfsson provided a bottom-line example of how such methodologies can result in increased profitability, noting that “Yahoo only makes 16% as much per page served as Google with the same underlying technology. Why? Scale.”

Brynjolfsson suggested to the CIOs in the audience that they push for experiments, measure and validate them in order to rapidly adopt the innovation, replicate it and then scale it. “Experiments aren’t an excuse to validate preconceived notions,” Brynjolffsson was quick to note. “That’s the wrong mentality. Leaders must approach experimenting with a genuinely open mind to see what works and what doesn’t.” Brynjolffson offered a CVS case study that he and Harvard Professor Andrew McAfee wrote in 2005 as an example. CVS created a pharmacy service improvement at one test location. Once the new process proved effective, CVS embedded the process into all of its IT system, replicating it to thousands of other locations.

Take his research with McAfee as another example. McAfee hypothesized that companies would become more similar over time as each organization enjoyed the benefits of improvements in information technology. What he and Brynjolfsson found was striking. When you compare leaders with laggards, over the past decade there has been a substantial growth in the gap. From the 1960s through the late 1990s, technology advancements benefited the nation’s companies in roughly the same amount.  Starting in the late 90s, however, there was a discernible shift to higher profitability in the top 25% of the nation’s corporations, particularly in more IT-intensive areas of economy. Increasing performance heterogeneity was a result that appeared to be closely correlated with IT - if not IT itself.

In other words , the results implied that companies were using information technology in a new way after the dot-com bubble, with the top echelon leveraging investments in ways that dramatically accelerated their growth and profitability in the new millennium.

When asked what CIOs and CEOs could invest in now for returns on investment in recessionary times, Brynjolfsson focused on so-called “Enterprise 2.0” technologies. In his view, blogs, wikis and enterprise microblogging quickly allow innovations to be discovered and amplified across the companies.