Last night Oracle announced their SPARC roadmap/strategy and a new so-called SuperCluster capable to run more than 30M TpmC! The event was hosted by Larry Ellison (Oracle’s CEO) and Jon Fowler (executive VP systems).
about the announcement
Hmmm! We all know, tpm-C is a very stupid, old and non-realistic benchmark but it’s still used to prove the theoretical horse power of DB machines, and yesterday we saw a gloating Larry showing the slide above: a “poor” 4M from HP, a “good enough” 10M coming from IBM and an “outstanding” 30M transaction per minute from Oracle!
It’s a shiny number but the result was achieved with a very particular configuration (Tons of very low computational power cores, Tons of RAM and Flash), I’m very curious about the behavior of similar configurations in the real world.
There are other supercluster configurations based on M series servers too, but benchmarks aren’t available at the moment and, finally, the SPARC-T3 powered exalogic cluster was also announced (GA in Q1-2011).
A small part of the announcement was about CPUs (not a lot indeed). One available now on M series server: the SuperSPARC VII+ (I think nothing more than a mere speed bump) and one available in a future (now in the Sun Labs) the T4 (a new 8 core CPU). Fowler also showed a very long (up to 2015) thin (a little bit built on air) roadmap with hardware and Solaris OS updates.
The announcement covered also an interesting new approach to some “pre-tested and optimized configurations”: you can ask Oracle for a better support service that covers your configurations to be sure that every single aspect of your environment will be followed and supported at the highest level possible. But only if your configuration is one among the few they name.
I’m just wondering the costs of this support options, sure it is not for everyone!
the mainframe style solutions and strategy
Oracle “mainframes” solutions sales strategy is on its way: certified lock-in bundles with high level support agreements as well as optimized with certified, secure and high performance characteristics! I’m not sure this is a good approach (at least for the majority of the customers) but it can work for some customers.
And… Oracle strategy will more likely succeed if the competitors won’t be able to compete! Yesterday I saw another, not very publicized, change of the famous Oracle processor core factor table. From now on, the multiplicator factor for HP Itanium cores will be 1 (it was 0.5)! It means that if you buy Oracle licenses for an HP machine you will pay an higher price if compared to the Oracle’s hardware (0.75 per core)!!! (I think that Oracle is the only enterprise DB that runs on HP/UX)
Now Oracle, when competing with HP (the weakest of the competitors), can hide the premium price for its overpriced and locked bundled solution. Who is going to buy an HP+Oracle DB server when you can have a fully integrated, better supported and fully optimized solution at a comparable price?
Oracle may kill each competitor with the multiplicator factor!