Big Data seems like it's definitely IT terrain - but not so according to a recent panel on Hadoop and Big Data. Instead, Big Data is coming into companies via the lines of business, with units funding individual Big Data projects, 70 percent of which focus on customer-facing ventures, such as boosting revenue or driving sales, according to a recent ZDNet article.
That matters for many reasons, not the least of which is it might impact how well big enterprise IT vendors fare in the Big Data market. Analyst Peter Goldmacher of the financial services firm Cowen & Company predicts this could be a niche where startups play better than "incumbent" vendors, such as IBM and Oracle.
Then again, IBM and Oracle don't usually focus on early adopters, but they often acquire startups to conquer new markets. So, I wouldn't bet everything against them just yet, particularly since Oracle just made its move in Big Data this week.
It's an interesting Big Data play for several reasons.
First, because it uses industry standards, you really can run Hadoop on any old servers you have sitting around unused - that's one reason it's so popular. That raises questions for analyst Curt Monash about whether that price is worth paying, even if it is packaged in a nice, integrated appliance. He told TechTarget there are downsides to a Big Data appliance, such as "overpaying for the same thing you could get much more cheaply elsewhere," and "losing out on some flexibility because you're stuck in a fixed appliance format."
That's one opinion. Brian Proffitt at IT World disagrees, saying Oracle's Big Data Appliance is actually a bargain when you calculate building a comparable system. He estimates to build a cluster with the same power, you'd need 385 nodes. At $4,000 a node - the typical commoditized Hadoop system price - he calculates it would take approximately $1.54 million to build a similar system. That's three times Oracle's asking price.
So Proffitt's good on the price. What he finds intriguing is the partnership with Cloudera:
The media reports suggest Oracle weighed creating their own Hadoop distribution, and didn't - a situation not unprecedented in Oracle's history. Students of Linux history will well remember that's exactly what happened when Oracle partnered with Red Hat to introduce commoditized Oracle offerings and then Larry Ellison and crew decided to roll their own Oracle Enterprise Linux in 2006 when they decided to cut Red Hat out of the stack.
It's like a dark cloud (ahem) looming over this partnership. " big data market is too big for Oracle not to want to own," Proffitt writes rather ominously.
Another interesting data point: Oracle is late to the ball on Big Data, says Pund-IT principal analyst Charles King. Heavyweights IBM and EMC entered via acquisitions - IBM with Netezza and EMC with Greenplum. While there's speculation that Oracle may acquire Cloudera, clearly that's not what's happening right now. Oracle underestimated Big Data, and now it's playing catch-up.
"Oracle kind of pooh-poohed the Big Data market for a while as its competitors moved forward more aggressively," King told CIO Today. "We've seen the company do something of an about-face over the past six or eight months."
It's also an intriguing move because, as CIO Today further points out, it flies in the face of a recent anti-Intel proclamation by Larry Ellison. The appliance uses x86 servers, which is what Hadoop typically runs on. That kind of means they're stuck unless, he adds, "Oracle can come up with something that can run Hadoop on SPARC."
Or, Proffitt might add, it builds its own version of Hadoop.