How does Oracle Prefetch work?

Question

There are 1000 rows in a table and I fire a "select *" query with a prefetch count of 100. How are the chunks of rows retrieved?
Scenario one: Get first 100 row, wait till cursor rolls over to 101th row, then fetch 100 more rows.
Scenario two: Get first 100 row, irrespective of whether this fetched rows are used or not, fetch the next 100 rows.
I thinks it is scenario one. Any idea? Any links to such documentation.
~Sri

Answer 1

I believe it works more like your first option.
Prefetch tells the OCI layer to grab x rows at a time whenever it has to grab more rows, and to buffer those rows on the client side. When your application requests rows, it can select 1 row at a time (for simplicity of program logic) and get that row from the cache OCI has on the client, rather than incurring the overhead of going to the database for each row. If your application requests rows that have already been fetched by OCI, it will get those rows. If your application requests rows that OCI has not already fetched, OCI will grab x rows from the server and then fulfill your request.
I'm assuming in the above discussion that your application is fetching fewer than x (prefetch size) records, since that's the case 99% of the time. I would assume that if your application fethes 2x records, that would be done in a single database round-trip rather than 2 round-trips, but I'm not sure of that.
Justin