I know very little, but do know that there is at least one specialized db server product that is set up specifically to mine such quickly-changing data very efficiently (as opposed to MySQL which was originally designed to mine infrequently changing data quickly, or a generalized RDBMS which is general purpose and designed to address a wide range of situations more or less inefficiently or efficiently depending on your POV).
And I could swear there was a reference here. If not here, then probably the PG hackers list.
Essentially the technique seemed to be to organize data not by tables but by denormalization into in-RAM structures based on knowledge of the types of queries that would be requested.
Not at all general purpose in the way an RDBMS is meant to be, and certainly not a replacement paradigm, but rather targetted for explicit needs, in this case fast queries against huge amounts of financial/market data.
This doesn't help much but at least perhaps will whet your appetite for searching further, both here and elsewhere.
Nice to hear from you, Jerry ... sounds like you're involved with some fairly heavy shit!
