Feeds

Google boasts of melting data center antidote

No chillers? No problem

7 Elements of Radically Simple OS Migration

Structure 09 Google is developing some sort of back-end technology that automatically - and nearly instantly - redistributes live compute loads when a data center is in danger of overheating. Or maybe this is just talk. Google prefers to at least maintain the illusion of data-center nirvana.

During a panel discussion last week at Structure 09, an annual mini-conference dedicated to all-things cloudy, Google senior manager of engineering and architecture Vijay Gill had a bit of fun with Google's well-known penchant for corporate secrecy - almost as much fun as he had at the expense of Microsoft.

When Facebook vice president of operations Jonathan Heiliger asked Gill to describe Google's now famous back-end infrastructure, Gill couldn't help but respond with a deadpan "Unfortunately, I can’t talk about that at all."

Gill promptly admitted this was a joke, and his audience chuckled. "Everybody at Google loves to say that," grumbled Gill's Facebook counterpart. But much the same words would resurface later in the conversation.

After Gill turned his nose up at Microsoft's back-end, Heiliger asked the sort of question that so often turns up during these conference panel discussions: "If you could wave your magic wand," Heiliger asked, "to create some bit of kit or new technology [for infrastructure] that we don't have today, what would it be?"

"Thanks for setting that one up, John," was the sarcastic reply from the irrepressible Gill. "We didn't discuss that one." But he semi-answered the question anyway, returning to the subject of Google's data center design, which he had at least alluded to earlier in pointing his audience to a recently-published paper on the subject: The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines.

"What we are building here - and of course this is the title of the [paper] - is warehouse-sized compute platforms," Gill said. "You have to have integration with everything right from the chillers down all the way to the CPU. Sometimes, there's a temperature excursion, and you might want to do a quick load-shedding - a quick load-shedding to prevent a temperature excursion because, hey, you have a data center with no chillers. You want to move some load off. You want to cut some CPUs and some of the processes in RAM."

Apparently, Google is working to detect temperature spikes and automatically respond to them in near real-time. Gill even seems to be saying that the company hopes to instantly redistribute workloads between data centers.

"How do you manage the system and optimize it on a global-level? That is the interesting part," Gill continued. "What we’ve got here [with Google] is massive - like hundreds of thousands of variable linear programming problems that need to run in quasi-real-time. When the temperature starts to excurse in a data center, you don’t have the luxury to sitting around for a half an hour...You have on the order of seconds."

Heiliger asked Gill if this was a technology Google is using today. "I could not possibly comment on that," Gill replied, drawing more laughter from his audience. Another panel member - LinkedIn vp of technical operations Lloyd Taylor, who was once director of global operations for Google - raised his hand as if to answer the question. But naturally, he didn't.

It's no secret that Google's distributed file system, GFS, and its distributed compute platform, MapReduce, are designed to deal with frequent hardware failures. When one machine goes down, its tasks are quickly picked up by another. But Gill indicated that Google is working to automatically respond not just to outright failures but to temperature spikes. He even seemed to be saying that the company is developing a way to sidestep the imminent breakdown of an entire data center ("a data center with no chillers") and to do so on "a global-level."

If Google has indeed developed a technology capable of such responses, it hasn't perfected the thing. If you've used Gmail in the past few months, chances are you're aware of this.

Google's strict code of corporate secrecy is often balanced by a pathological need to tell the world how great it is. And Gill showed a bit of both last week. It was entertaining - if not enlightening. ®

Best practices for enterprise data

More from The Register

next story
Sysadmin Day 2014: Quick, there's still time to get the beers in
He walked over the broken glass, killed the thugs... and er... reconnected the cables*
VMware builds product executables on 50 Mac Minis
And goes to the Genius Bar for support
Multipath TCP speeds up the internet so much that security breaks
Black Hat research says proposed protocol will bork network probes, flummox firewalls
Auntie remains MYSTIFIED by that weekend BBC iPlayer and website outage
Still doing 'forensics' on the caching layer – Beeb digi wonk
Microsoft's Euro cloud darkens: US FEDS can dig into foreign servers
They're not emails, they're business records, says court
Microsoft says 'weird things' can happen during Windows Server 2003 migrations
Fix coming for bug that makes Kerberos croak when you run two domain controllers
Cisco says network virtualisation won't pay off everywhere
Another sign of strain in the Borg/VMware relationship?
prev story

Whitepapers

7 Elements of Radically Simple OS Migration
Avoid the typical headaches of OS migration during your next project by learning about 7 elements of radically simple OS migration.
Implementing global e-invoicing with guaranteed legal certainty
Explaining the role local tax compliance plays in successful supply chain management and e-business and how leading global brands are addressing this.
Consolidation: The Foundation for IT Business Transformation
In this whitepaper learn how effective consolidation of IT and business resources can enable multiple, meaningful business benefits.
Solving today's distributed Big Data backup challenges
Enable IT efficiency and allow a firm to access and reuse corporate information for competitive advantage, ultimately changing business outcomes.
A new approach to endpoint data protection
What is the best way to ensure comprehensive visibility, management, and control of information on both company-owned and employee-owned devices?