“Historically, RDB's are dying, simply because that are too rigid to evolve into fragmented, globally distributed,
highly replicated file systems. Flat is Back.”
Throwing the baby out with the bath water, I’d say. The relational aspect of a modern RDBMS database product is hardly the whole product. Query optimization can be done within
a multi-field query of a single table, for example. Indexing is valuable for any database, unless there is really good accelerator technology available. ACID transaction properties are also important in many use cases. A lot of these data warehousing or
NoSQL systems are comparatively immature implementations compared to DB2 or Oracle or Postgres. Sharding is straightforward to layer on to a RDBMs from the client side, that’s hardly a reason to switch to another technology.
What is the problem with stored procedures? Clearly in any client/server architecture there will be situations where code needs to be close to the data for performance reasons.
Anyway, this has nothing to do with the value of statistical inference so far as I can tell.
Marcus
| Free forum by Nabble | Edit this page |