|< < 24 > >|

File Organization

Primary Tree Index

Range search

The wider the range, the more random accesses will be required.

"Folklore" (from the 70s) says that if the range covers more than 10% of the records, it's better to do a sequential file scan. (I'm not sure that's still true.)

Possible alternative:

• Scan the range in the index, collected the record ids.

• Sort by page address.

• All accesses to a page will be consolidated.

|< < 24 > >|