Is the size of an extent predictable?

Hello all!

I’ve got a question related to an older discussion.

It seems to me like the size of an extent is not predictable. Example: We got one ADABAS-file with a bigger growing rate. So it’s normal, that ADABAS builds new extents for NI automatically. Sometimes the size of these extends are about 7000 blocks. The next extent is created with roughthly 20000 blocks.

My question is: Why? The truth is: ADABAS didn’t fill gaps in the database-layout with that. How does Adabas decide, how big the next extend will be?



Hello Matthias,

Here the automatic extent allocation algorithm for v3.x:
For the Address converter generally 12.5% of the already allocated space is allocated.
For the other extents the size of the new extent is:
((MAXISN - TOPISN) / TOPISN) * size of the already allocated space, but at least 12.5% and not more than 50%.
Note that this is the current formula used in v3.x, which may be changed in other versions.