Larry Ellison is famous for shaking up the IT industry. Now at the opening to Oracle OpenWorld 2013 he seems to have done it again with his announcement of In-Memory analysis for Oracle’s Oracle 10c database. Readers can see live analysis and discussion of this and other announcements live from TheCUBE on the floor of Oracle OpenWorld through Wednesday.
In-Memory is an obvious belated response to SAP’s HANA, and a comment from SAP is unsurprising. However, SiliconAngle has heard from new-SQL startup Clustrix, which designed its database for horizontal scale-out and exploitation of flash, and surprisingly, from IBM, which as a matter of policy almost never comments on its competition.
Clustrix CEO Robin Purohit argues that in this day of Big Data and the growing importance of Open Systems, the old vertical build-up stack that Oracle is defending is expensive, inflexible, and filled with proprietary lock-in. The Open Systems build-out model, in contrast, allows users to closely integrate whatever technologies they need — for instance the Clustrix database to support transactional processing and near real-time analysis of that transactional data combined with Hadoop and possibly a NoSQL database to capture unstructured data associated with those transactions for in depth predictive analysis of the market.
“Our fundamental viewpoint is that the world is transitioning from scale-up to scale-out, and it’s just a matter of time before the enterprise shifts all their databases to scale-out,” he says. “Oracle is offering a scale-up system that is very expensive and does not provide the near real-time analysis of transactional data that our customers want.”
Clustrix uses an architecture that writes all transactions in memory in parallel across the cluster, writes the transaction log to a non-volatile RAM card, then writes it to flash memory on the server to protect against a power failure and from there to an array for long term storage. Analysis can be conducted on the data in memory as soon as it is captured. It offers its database technology either in-house, pre-installed on a server, or in the Amazon IaaS Cloud.
IBM offers both build up and, through its PureData for Transactions and DB2 PureScale clustered systems, a build-out option, says IBM VP of Databases and Data Warehousing Sean Poulley. “We are driven not by technology but by the customer’s needs, and we can provide either option. And we can provide them now, while we have yet to hear a delivery date from Oracle.”
IBM DB2 can deliver a single database system built on PureSystems that supports both transaction processing and near real-time analysis, leveraging large amounts of flash, he says. IBM has integrated DB2 down to the thread level on the processor chip on its PureSystem servers to get maximum performance. This, he says, is why it gets higher performance on IBM’s Power RISC chip than on x86 — Power can handle more threading.
One of the weaknesses of the HANA and Oracle In-Memory systems is that if the system suffers even a momentary power loss it loses all the data in memory, and it has to be reloaded. IBM’s approach of writing directly to flash memory in the server protects the data from power outages and other glitches while providing near real-time data analysis due to the level of integration between the software and hardware. With DB2 IBM can provide a recovery point of 13 milliseconds.
DB2 automatically captures data in flash on the server as it comes in and maintains all actionable data in the flash cache, while copying it to the array in background. Also, IBM Actionable Compression allows analysis of compressed data without decompressing. This doubles or triples the effective capacity of the flash cache.
One of Oracle’s big issues in the last year has been that it is not growing. IBM does not break out its figures for specific products such as DB2, but Poulley says it is seeing significant market growth for both the database and underlying hardware, which is driving IBM R&D investment in DB2. Obviously this is working off a much smaller base than Oracle, which dominates the RDBMS market. Exactly what this means for Oracle’s future, and whether In-Memory will drive growth of Oracle database systems and of Sun hardware as Oracle hopes in an increasingly competitive market, remains to be seen. However, obviously Oracle faces new challenges from both Open Source, NoSQL and New-SQL database vendors such as Clustrix, and from IBM.