[Mix published from quant.SE]

I am searching to transform some tickdata .csv files right into a database on the local disk after which use R to call the information and do my various statistics and modelling.

What exactly are some guidelines / implementation techniques that may be suggested to reduce any head aches?

Noted that packages for example mmap help enormously, but have considered trying and discover a far more 'permanent' solution.

The best idea db to experience with? MySQL appears to become suboptimal as it is relational (instead of column-oriented). q/KDB appear to possess a trial version to experience with but includes a very steep learning curve. The best idea db to deal with and offer tick demands?

Any help much appreciated. Platform agnostic, however i suppose Linux is my preferred platform.