I have been writing an app in order to archive all my media (primarily MP3's) into a SQL server database on another machine. then to return them as quickly as possible using various search techniques

i have written up to a complete database

process being

1.) build up string[] of files using directory.getFiles
2.) build up object MP3 using 3rd party Taglib dll by passing string
3.) create List<MP3> from each
4.) passs List<MP3> to SQL


this works brilliantly and i run it multithreaded in order to stop not responding problems

now the next part im going to create a UI in order to search and launch each file (much like a media centre) however i am primarily concerned with the speed of search.

i will be searching for artist, genre, album and track name

some specs

mp3 table contains
ID , filename , extension , pathstring , Artist , Album , Track , Genre , artistID , AlbumID , Genre ID

then have
Genre , Artist , Album
tables

containing
ID Description

that links to MP3 table

MP3 table is about 15000 rows Long

now because i'm primarily concerned with speed i was wondering

is it best to run straight against SQL (bearing in mind its on another box that is about 6 years old) or would it be best to build some lists on app load (memory usage is not an issue) therefore searching against the lists and not against SQL