Testing times for Oracle’s in-memory database option
Software giant under fire for buggy software and stealthy default charges.
Oracle has had a rocky run of it with its much vaunted Oracle Database In-Memory option, first announced at JavaOne 2013. Admittedly, in-memory was always going to be a hard-sell for a company whose CEO had publicly dismissed the technology as “wacko.” Nevertheless, Larry Ellison bit the bullet and appeared in person to launch the technology this June, peppering his speech with a series of widely derided claims of “ungodly” speed and unique functionality.
Of course, rivals SAP Hana, who’ve been doing in-memory since way back when (for Ellison’s sake, hopefully far back enough that nobody remembers his comments that the SAP crew were “on drugs” if they believed it made them true contenders for the DB throne) were quick to jump in, arguing that Oracle were merely playing catch-up, and along the line customers would be stung by sneaky hardware costs.
Plucky startup Hazelcast also joined the fray, choosing the very same launch date for Hazelcast Enterprise, a commercial offering from which will go head to head with the Java overlords. Both systems have scalability, high availability, reliability, cloud readiness, convergence of the OLAP and OLTP approaches, and compatibility with SQL in common, but where their paths diverge is that Hazelcast are all about low-cost, open source software, whilst Oracle operate on a path geared towards vendor lock-in, leading Hazelcast to label the San Francisco behemoth an Apple wannabe.
Aaron Auld, CEO of in-memory database firm Exasol, jumped in with gloomy predictions that Oracle’s new feature would be ‘unfocused, hardware-centric and expensive’ - and it’s looking like he may have had a point. This week, it’s emerged that the in-memory option is switched on “by default” in Oracle Database 18.104.22.168, raising the hackles of clear-licensing advocates. For use of this feature, hapless customers will apparently be charged a minimum of $23,000 (£14,000) per processor used by the database on Oracle’s SPARC servers.
Oh, and there’s also the matter of a little bug, which after much to-ing and fro-ing on various platforms, Oracle has finally had to cop to. Although it remains unclear if the little blighter poses any real risk, it does cause the in-memory database option to be reported as being in use when it’s not, which is pretty annoying. Ellison’s patch patrol will issue a fix for this around October time. Let’s hope nobody gets stuck with a trifling $23k or so tab as a result of this little flaw before then.