I'm currently running postgresql 7.4.8. Our two biggest tables have
indexes on (x, y), where x, y are integers, and there are often
many y values per x value. The ratio can be anywhere from 1:1 to 200000:1.
In the 8.0 release notes, (section E.10.4.1), I noticed this
statement:
Improve B-tree index performance for duplicate keys (Dmitry Tkach, Tom)
This improves the way indexes are scanned when many duplicate
values exist in the index.
Can someone describe these improvements in more detail? How were such
scans done in 7.x, and what exactly is different in 8.x?
Jack Orenstein