Back in April, Steve Miller suggested that approximate BI could be a growing trend, gaining speed at the expense of (often false anyway) precision. That idea of course goes well with Infobright’s recent released Rough Query feature, and also with Datameer’s year-earlier pitch.
Aster Data (now a Teradata company) is positioning itself as analyzing multi-structured data — which is my second-choice term, behind the more precise but odder-sounding “poly-structured.” I hope “poly-structured” wins, and plan to keep using it myself; but I recognize that “multi-structured” may actually be the one that prevails.
Barbara Darrow wrote a great piece on Oracle Exadata pricing. Highlights include:
- Routine Oracle software discounts are high.
- Exadata discounts are higher.
- Big/referenceable customers get the best Exadata discounts. The term “extremely deep” was used. (I’ve also heard that from Oracle competitors, with the term “free” even coming up, hyperbolically or otherwise.)
- Oracle’s hardware maintenance pricing is forcing users to trash Sun gear, even when it’s working. One guy told the story of literally crying as the Sun boxes were taken away.
- Oracle’s 22% of license maintenance fee goes up to 27% after two years. I didn’t know that.
Oracle has been making considerable messaging fuss around a win in Japan, where Softbank replaced years-old Teradata systems with vastly less new Exadata gear. I blogged that this is hardly an apples-to-apples comparison. During my visit last April, Oracle pushed back, in particular pointing out that the Softbank division that awarded the deal was very separate from the one that was an Oracle reseller. But Monday Teradata shared with me a counter-pushback, asserting that during the recent worldwide recession, Softbank assigned its underemployed systems integration division to do internal projects — including the data warehouse upgrade. I.e., Teradata stands by its claim that this replacement was strongly influenced by the Softbank/Oracle partnership.
If you’re analytically inclined, Kx Systems has some interesting ideas, manifested in kdb+ and so on. A 2009 ACM article seems as good a starting point as any, the company’s website probably aside. Confusingly, Kx is small company that evidently does most of its selling through a couple of much larger partners. Also, 1010data happens to be built on an older version of Kx’s technology.