Google search primed for 'Caffeine' injection
A shot in the back-end
Google's overarching philosophy is to build a single, distributed architecture that runs all its services. And Cutts acknowledged that many of the back-end tools that drive the new indexing system - including GFS2 - will eventually be put to use across other Google services.
Part of the appeal of GFS2 is that it's specifically designed to handle low-latency applications, including Gmail and YouTube. With the original GFS, a master node oversees data spread across a series of distributed "chunkservers." For apps that require low latency, that lone master is a problem.
"One GFS shortcoming that this immediately exposed had to do with the original single-master design," former GFS tech lead Sean Quinlan has said. "A single point of failure may not have been a disaster for batch-oriented applications, but it was certainly unacceptable for latency-sensitive applications, such as video serving."
GF2S uses not only distributed slaves, but distributed masters as well.
In recent weeks, Mountain View has also acknowledged the existence of a new back-end technology known as Google Spanner, a means of automatically moving and replicating loads between the company's mega data centers when traffic and hardware issues arise. But a company spokesman tells us this is not part of Caffeine, although he says that "both [are] part of an ongoing company-wide effort to improve our infrastructure."
In a recent presentation (PDF) at a distributed-computing shindig in Montana, Google fellow Jeff Dean seemed to describe Spanner in the present tense. Though he declined to discuss the presentation with The Reg, he indicated that all the information in our recent piece on the mystery technology is correct.
According to Dean, Google intends on scaling Spanner to between one million and 10 million servers, encompassing 10 trillion directories and a quintillion bytes of storage. And all this would be spread across “100s to 1000s” of facilities across the globe.
Today, Google operates roughly 40 data centers, and it seems that Caffeine will be deployed one facility at a time. According to Cutts, this involves taking each data center offline and shifting its load elsewhere.
"At any point, we have the ability to take one data center out of the rotation, if we wanted to swap out power components or different hardware - or change the software," he said. "So you can imagine building an index at one of the data centers and then copying that data throughout all the other data centers. If you want to deploy new software, you could take one of the data centers out of the traditional rotation."
So, somewhere in the world, there's a mega data center on the verge of sabbatical. Or perhaps it's already happened. Presumably, Google will tell us at some point. And tell us very little. ®
Sponsored: Hyper-scale data management