What is the performance degradation for huge partitioned tables?

[Per Arne Engstad]

For a project we want to store up to 5 billion (5.000 million) rows in a single database table. In order for this to work we need partitioning, and we probably need to use hash partitioning on a text field (URL). The table has a couple of indexes; a primary key (string) and at least one integer column which contains an index.

Do you have any performance numbers/graphs which show:
- the decrease in insert time per row as a function of number of rows in the table
- the increase in select time per unique row as a function of number of rows in the table

Do you have any experience wrt. maximum number of rows using partitioning in real applications on single-CPU Linux hardware?

I hope you can help.

Best regards
Per Arne