SAP has been working on in-memory database functionality for a while, and has been running its Business Intelligence Accelerator (BIA) using an in-memory database for a couple of years. But at this Mays Sapphire user conference, SAP made a point of showing how to make mincemeat out of the throughput of traditional relational databases.
With SAPs in-memory system providing sub-second response rates against a billion records, SAPs Hasso Plattner vowed that the company was close to being able to run a Wal-mart-class database weighing in at a petabyte or two completely in-memory and with the same blinding response time.
But thats not all, and heres where the revolution comes in. Plattner finally made formal whats been stewing at SAP for years: this same technology is slated to be deployed in the traditional world of SAP Business Suite transactions, promising more than just speedy throughput.
Of course, the news that SAP was planning to replace its relational database backend with a column-based in-memory system would also make things a little complicated with IBM and Microsoft, both of which have been pushed by SAP to challenge Oracles dominance, with some success. So Plattner stuck to more of a techy exposition of what the technology can do for customers than a marketing exposition of what it could to the competition.
But its no secret who would have the most to lose from an in-memory revolution. At least until Oracle mainstreams the in-memory technology it bought with the acquisition of TimesTen in 2005.
Because once SAP can truly provide an in-memory alternative to Oracle, it will have a massive performance benefit that Oracle simply wont be able to challenge, thanks to Oracles closed database design. While SAP has always been database agnostic, Oracles applications are tied directly to its database, and therefore cant be run on an in-memory database, theirs or anyone elses. Oracle could and probably is already thinking of providing some version of in-memory support for its forthcoming Fusion Applications suite. But thats not what it will take to blunt an SAP in-memory drive against Oracles purely relational eBusiness Suite.
There are other benefits to in-memory as well that accrue to customers, and not just to SAPs competitiveness: in-memory is cheaper to buy, obviously doesnt need the massive server infrastructure that a relational database needs, and, at least theoretically, is much cheaper to run as well, as in no corps of expensive DBAs needed.
It would be foolish to predict that the relational database will go the way of the dinosaurs any time soon, we know from years of predicting the total demise of technology that most functional systems survive in some form or another, despite their growing irrelevance and obsolescence. (Do a search on IBMs IMS and youll see how popular that old warrior still is after 40 years in service.)
But just as most useful technologies seem to persist, all eventually succumb to challenge from something better, faster, and cheaper. This ascendancy of in-memory databases and the concomitant decline of relational database plays exactly to this trend, and SAPs Plattner very neatly defined how far along relational databases are in the decline category. Not finished, not out of the market yet, but the technology guillotine is being sharpened, and relational databases are about to be cut down to size. With SAP cheering all the way.