[GENERAL] vacuum taking an unusually long time

Vacuum operations on several tables are taking much longer than they previously were.

We currently have 3 autovacuum processes that have been running more than 3 days each.

The tables are large (between 40 and 90GB each).

Postgresql version is 8.3.1

maintenance_work_mem is 512MB  (on a 32GB server).

Any ideas what would make vacuum take so long?

What can I do to speed things up?

thanks,

- Mason