Finally it seems Oracle’s end-to-end solution for Big Data is getting clear. At this week’s Oracle Open World, I saw the webcast of one presentation on the subject. The hardware-software combo story (engineered system as they call it, after Sun acquisition) clouds other key software solution pieces. First of all, the Oracle 12c (c for cloud) in-memory (IM) database got a lot of attention at last Sunday’s kickoff by Larry. On the stage, there were two interesting demos, one for the ‘small’ processor scanning 7 billion records a second single thread. On the M7 the scan runs parallel at 341 billion records a second. There were no numbers for the M10 but it should even be faster. IM is heavily based on columnar technology, but columnar is used as index and not as storage. Interesting stuff. It’s limitations will come clear as we know further details.
On the Big Data, the traditional structured data store via Oracle DBMS on the engineered system Exadata is old story. They also have Hadoop platform (partnership with Cloudera) plus a NoSQL data store with JSON, for unstructured or semi-structured data. They can stream Hadoop data to Oracle DBMS via HIVE. For analytics, they have SQL tools and also R language support for Hadoop Map-Reduce operations. They also announced Oracle Fast Data, for dealing with “data in motion” such as sensor data. The customer examples (NY Stock Exchange, Airbus,..) all seem to use the Exadata and Exalytics appliances and of course they will benefit with the in-memory option of 12c.
The Oracle 12c in-memory columnar data store is clearly aimed at combating the SAP Hana onslaught. The Big Data story is still hard to piece together, unlike the clarity of message from IBM which has a dedicated division for Big Data. The actual use of the NoSQL or Hadoop solutions are not evident yet, as the sales folks must be pushing the Exadata appliance offerings for a bigger commission. However, for the first time, I saw an end-to-end solution with some clarity.