Kbase P164449: What are some reasons that a database size on disk may be significantly smaller after a dump and loa
| Autor |
Progress Software Corporation - Progress |
| Acesso |
Público |
| Publicação |
4/22/2010 |
|
Status: Unverified
GOAL:
What are some reasons that a database size on disk may be significantly smaller after a dump and load?
GOAL:
Can a database be smaller after an dump and load of the database.
FACT(s) (Environment):
All Supported Operating Systems
Products / Versions
FIX:
The size difference on disk that you see between the database before a D&L process and the database after a D&L process can be attributed to many factors.
You can see a tremendous reduction of the size of the database on disk after a dump and load even though the record count has not changed.
The database structure related to fixed extents can be changed to account for less fixed space on disk then was held fixed before the dump and load as long as the space was not actually needed and the new structure has more than enough space to hold the actual data/indexes . That can show up as a large drop in database size on disk.
Another big chuck could be accounted for by the compaction of indexes so that fewer blocks are used than before. That portion of space will eventually be used up again as the indexes grow and split with adds/changes being done to the data. looking at the dbanalys of the indexes, the blocks used and the utilization percentage you will see that after the compaction in the D&L process there can be a large drop in the number of blocks used to store an index. You would need to go through all the tables in an analysis of the indexes prior to the D&L as compared to after the D&L to add up the reduction in blocks due to index compaction.
Another factor is the white space left in the database due to deletions of records. If there were large purges of data the space could be huge. All that space will recaptured during a dump and load.
If tables were moved out of the schema area without running the move schema utility, that space wouldl be recaptured as well.
Changing the records/block and the cluster size could account for additional space changes required on disk.