Kbase P85904: Index area and records per block considerations
| Autor |
Progress Software Corporation - Progress |
| Acesso |
Público |
| Publicação |
6/21/2006 |
|
Status: Unverified
GOAL:
Index area and records per block considerations
GOAL:
Should you set your index areas records per block to 1
FIX:
Index blocks do not contain records.
If a storage are contained only indexes, then it contains no records.
Index blocks are identified by solely are number and block number. The storage manager uses an identifier we call a "dbkey" to fetch blocks. It is like a rowid, except that the record number within the block field is all zeros and is not used.
By setting the record-per-block to 1, you minimize the size of the "record within block" part. That allows you to address more blocks than 64 gbyte. Which probably does not matter in most situations. A 64 gbyte index is extremely large.
It has been reported by several people that there is also a small performance advantage to using 1 for index areas. Index blocks above the leaf level of the tree will compress a tiny bit better.