EMC Positions Big Data as a Platform for Business
One of EMC's big data announcements this week at EMC World was its upcoming release of its own Apache Hadoop distro. This is a big deal for enterprise customers, particularly those inching closer toward membership in the Petabyte Club (EMC has over 1,000). Most are quickly realizing their massive amounts of unstructured data simply can't be analyzed in traditional relational databases like Oracle.
EMC's strategy takes a page from Microsoft's infamous playbook on "Embrace and Extend". EMC intends to add innovations like replication, 5-9s availability, and management which are all capabilities lacking in the Hadoop Distributed File System (HDFS) today. Furthermore, these are features that mainstream enterprises expect from analytics software frameworks, particularly if data is to be a platform for business -- from which to extract value and make money.
One of the most intriguing innovations previewed was what EMC referred to as Pluggable I/O. This software layer essentially gives customers the choice of using standard HDFS or other file systems like EMC's Isilon OneFS, Symmetrix File System (SFS), pNFS, etc. to store Hadoop data.
It's more apparent where EMC is going with this strategy: Acquire companies like Greenplum and Isilon that are at the center of this new era of big data, leverage their critical technology assets and DNA, and integrate it into your existing storage IP that's got a huge installed base. Then you sell that same customer base even more kit to harness the value of all their structured + unstructured data that resides on EMC storage arrays today. Sheer brilliance!
EMC's bold move would certainly seem to put Oracle in a defensive position. To date, Oracle has been MIA on Apache Hadoop, choosing instead, to have a maniacal focus on Exadata. In Oracleland, Exadata can do it all!
Four months ago, a Merrill Lynch analyst questioned Oracle's SVP of Oracle Server Technologies, Andy Mendelsohn, at its annual financial analyst conference about the growing popularity of NoSQL technologies such as Hadoop. Mendehlson was downright dismissive.
Mendelsohn suggested the analyst's question gave rise to an interesting academic discussion. He went further with his dismissive retorts by saying that NoSQL deployments were mere research projects. Is he for real?
The fact is that while the world's largest financial services and telecommunications companies still have plenty of reason to use Oracle databases, they also can't solve their biggest business problems without Apache Hadoop and other NoSQL technologies like Cassandra. And every day, more companies are exploring new ways to integrate Hadoop with their Oracle relational databases -- often in ways that significantly cut Oracle database licensing and support spend.
Oracle's silence is deafening. EMC gets it, Teradata with its acquisition of Aster Data, gets it, HP Vertica gets it, and IBM, through its numerous partnerships, gets it.
The reality is the 1960's IBM T.J. Watson era of providing a closed-end IT solution that Larry Ellison so fondly intends on resurrecting, isn't coming back.
Fast forward to 2011, and there are simply too many choices, too much open source innovation at global scale and accessibility, and dramatically declining infrastructure costs that exploit commodity Intel multi-core processors and flash memory technologies. This wave is inevitable, and not even Oracle's Ellison, the King of relational databases, can stop it.
Oracle would do well to mitigate the growing Hadoop threat by acquiring Cloudera. It fits with Oracle's preferred buying over building strategy. Not only has Cloudera been working with Quest Software to integrate Hadoop with Oracle via the Ora-Oop connector, but Cloudera's CEO is the former chief of Sleepycat, the Berkeley DB open source database company Oracle acquired in 2006.
It's only a matter of time before an equally powerful player in the enterprise space swoops in to buy Cloudera. But in all likelihood, Exadata will continue as Oracle's big hammer where all markets and enterprise customer requirements are nails.
Facebook explains what happens to your data when you die - Facebook Inc. has more than 2 billion people using its social network, more than a quarter of the Earth’s population, and with such a large user base, it...