October 10, 2010
When vendors talk about the integration of advanced analytics into database technology, confusion tends to ensue. For example:
- Aster Data is generally an exception to this rule, as it should be, since that integration is at the core of its positioning. Even so, in the last paragraph of that link, I called Aster out for what at that time was some product description nonsense, which was specifically in an area that many vendors are confusing about explaining, namely …
- … the distinction between three kinds of parallelization.
- If you do something entirely in SQL on an MPP system that parallelizes SQL — then it’s parallel!
- If you have a parallelization framework such as SQL or MapReduce that can invoke the same function on every node — well, then that’s parallel!
- Many algorithms — including almost every important statistical one — have to be explicitly coded to be parallel if they’re actually going to run in in parallel. The seminal paper on parallel data mining shows that such parallelization is, in many important cases, straightforward — but somebody still has to take the trouble to actually do it.
- Netezza TwinFin i-Class was renamed/repackaged/repriced before it ever shipped. Even so, when Tim Young or Phil Francisco tries to recall exactly the “i” stands for, comedy ensues. And the post I promised to write about Netezza TwinFin i-Class in June (as per the last sentence of this post) hasn’t happened yet, for reasons other than lack of interest on my part.
- SAS/DBMS integration tends to be a multi-year process, with in-database scoring coming long before in-database modeling. The drip-drip-drip of big-company PR over that time period can be quite bewildering …
- … especially since SAS partners in some cases are shipping home-grown in-database modeling long before SAS gives it to them.
- After backing off from its early endorsement of MapReduce, Greenplum pretty much went to the other extreme and didn’t talk about its advanced analytics capabilities at all.
Categories: Aster Data, Greenplum, Netezza, Predictive modeling and advanced analytics, SAS Institute
Subscribe to our complete feed!