Though it will eventually offload search to Steve Ballmer and Microsoft, Yahoo! has no intention of shrinking the new back-end infrastructure now driving its search technologies.
"What impact does the Microsoft search deal have on the Yahoo! cloud? It has no impact," Shelton Shugar, Yahoo!'s senior vp of cloud computing, tells The Reg. "All the services we're building, we will continue to build. All the roadmaps we have in place, we will continue to work towards."
For the past few years, Yahoo! has worked to build a unified collection of back-end platforms that can run all its online applications. Or at least most of them. Among other things, this so-called "private cloud" relies heavily on Hadoop, the open-source number-crunching platform based on Google's proprietary infrastructure.
According to Shugar, the company's search technologies eat up less than half of the resources served by its cloud, and when search departs for Redmond, other services are sure to fill the hole. "Search is just one of the many services [in use across the platform]. We use it for advertising. We use it for content. We use it for all of our usage logs. It is essentially becoming the data warehouse for all of Yahoo!" he says.
The company now has about 25,000 servers running Hadoop, and Shugar says "search is not the majority user."
Following the July agreement with Microsoft, Hadoop founder Doug Cutting left Yahoo! for the Hadoop-happy startup Cloudera, raising questions about Yahoo!'s future with the open source project. But Shugar says the company may even grow its core Hadoop team. "There will be no backing away from Hadoop. You may even see an acceleration."
Yahoo! is Hadoop's largest contributor, and the company will join the community as it descends on midtown Manhattan later this week for Hadoop World, the platform's first east coast developer conference.
Shugar says Yahoo! is still at work on an SQL-like Hadoop programming language that dovetails with a lower-level language known as Pig. Appropriately enough, the Yahooligans call it Pig SQL.
Yes, there's already an SQL-like Hadoop language: Hive, developed at Facebook. But much to the Facebook's chagrin, Yahoo! feels the need to build a language that better suits its existing Hadoop infrastructure. "We have this Pig stack. We have a lot of things built on this Pig stack, including an optimizer," Shugar says. "We want to keep that stack streamlined."
That said, the new language will be open-sourced.
One thing Yahoo! will not do - at least for the foreseeable future - is open up its infrastructure to external developers, as Google, Microsoft, and Amazon have done. Its private cloud will remain private. But, Shugar points out, Yahoo! is juicing developers through its open source contributions. Despite the media-fueled notion that it's nothing but an online advertising outfit, he insists, Yahoo! is still a technology company. ®