Wolfgang Schölmberger
Member
Hi everybody,
OE11.7, Windows 8, 10
We have a Type-I-database that has only one table - it is used as an document-archive. So there are only creates, no updates at all and very few deletes.
Every document is written in chunks of 16K, so if a document has e.g. 100K it needs 6 records which are all filled up and the rest goes into one more record.
with 650.000 records tabanalys shows the following statistics:
Max recordsize: 16422
Average recordsize: 15373
Minimum recordsize: 23
I wonder if it would make sense to change it to a Type-II-database. Apart from the fact, that the indices of this table should be in an own area, I would guess: NO.
The database we are using now is a 4K database. We might benefit a bit if we changed to a 8K database, so that 1 record uses 2 blocks instead of 4 as it does now, but in the end I would think this is neglectable. And as there is only 1 table it shouldn’t be important, whether the database has one block per cluster or 512 blocks per cluster.
Of course there is the index-problem, but apart from this ?
TIA, Wolf
OE11.7, Windows 8, 10
We have a Type-I-database that has only one table - it is used as an document-archive. So there are only creates, no updates at all and very few deletes.
Every document is written in chunks of 16K, so if a document has e.g. 100K it needs 6 records which are all filled up and the rest goes into one more record.
with 650.000 records tabanalys shows the following statistics:
Max recordsize: 16422
Average recordsize: 15373
Minimum recordsize: 23
I wonder if it would make sense to change it to a Type-II-database. Apart from the fact, that the indices of this table should be in an own area, I would guess: NO.
The database we are using now is a 4K database. We might benefit a bit if we changed to a 8K database, so that 1 record uses 2 blocks instead of 4 as it does now, but in the end I would think this is neglectable. And as there is only 1 table it shouldn’t be important, whether the database has one block per cluster or 512 blocks per cluster.
Of course there is the index-problem, but apart from this ?
TIA, Wolf