Actually this originally was in a database, but the performance.... Benchmarking suggested it would take 5 months to complete, and since this was to be an annual task we're aiming to get it down to two or three weeks. Since DB access was the second biggest cost, it appeared to be an obvious place to tune, so this is an attempt to *replace* the DB look-ups and to a space/time trade-off.
| [reply] |
I can't say I've ever even tried to put that much data into memory at once, but I have dealt with some large collections. You might try chunking up the original data and storing it in temporary tables you can iterate through so you have less to work with at one time. If your database provides something like LOAD DATA INFILE, you can make use of it to dramatically reduce transaction times when it's time to put all the data back.
That all said, if it were mine to do the single word that would keep me awake at night is "cluster".
Good luck and have fun with it!
| [reply] |
