SAN JOSE, Calif. -- Intel wants to drive big-data analytics toward open-source software accelerated on its processors.
In a first step in that direction, it is working on an upgrade of its version of Hadoop that blends in features from the distribution provided by Cloudera, a leading open-source supplier of the code. Meanwhile it has already started working with customers to determine what sort of analytics apps they want on top of Hadoop and how to accelerate them on x86 chips.
Intel sees Hadoop as a sort of operating system for big data, "the freeway we have to put in place," says Ron Kasabian, Intel's general manager of big data, speaking in an interview with EE Times. But Hadoop has lacked enterprise-class features such as backup, recovery, access control, and security. Intel has tried to add such features in its distribution, but recognized the Cloudera code was more popular and had better streaming data and other features. Therefore it struck a deal to merge the code bases.
"We have a road map that says by the end of the year a confluence of the two products will be complete," says Kasabian, a former IT manager at Intel.
"Now we are turning our focus to analytics apps that run on top of Hadoop," he says. "We are doing proof-of-concept projects with customers to understand what analytics workloads look like … [in part] to figure out if we can do anything in hardware to accelerate them," he adds.
For example, emerging workloads for Internet of Things applications may have unique requirements. With this work Intel is seeking strategic opportunities for end-node CPUs such as Quark or cloud processors such as its Xeon. It is also looking for ways to involve its McAffe and Wind River groups in analytics.
Today "the analytics that work on Hadoop are a step or two behind," he says. "We are starting to see analytics from a lot of small companies come to market, but I'm not seeing a lot of open source in the analytics. My goal is to get to a point where there's broad availability of analytics in open source on Hadoop," he added.
— Rick Merritt, Silicon Valley Bureau Chief, EE Times