I would emphasise the point made about Postgres read performance. If all you are doing is reading single rows from a single table using an index, then yes Postgres is overkill and other solutions may perform better. However, the overhead that Postgres has is due to a remarkably effective query planner, which can be a godsend if you have a complex multi-table join. A few years ago I was working on a project with a large Postgres database that regularly did queries joining 20 tables together, which Postgres was quite happy to accommodate.

Also, that overhead can be suffered only once for many similar queries if you make use of prepared statements, which can bring the performance of simple read queries up.

Also, bulk writes can be sped up through the COPY IN BINARY interface - we were getting bulk load speeds basically limited by the write speed of our RAID array.

In short, I have been very impressed indeed with Postgres. They do it right.