Increasing performance when sequentially reading an indexed file
A customer has noticed that on a rather large indexed file, the performance seems to suffer when reading sequential through the entire file.
Is there a way to increase the performance of indexed file I-O when doing sequential reads?
They might look into using the extfh.cfg option called NODESIZE.
This controls the size of the index nodes that are used for a file when it is created.
The larger the value, the more keys that can be stored into a single node and this can speed up sequential processing quite a bit.
In a recent test when reading sequentially through an indexed file containing 8 million records, the following timings were achieved by setting the NODESIZE and recreating the file:
1024 3 mins 55 secs
4096 3 mins 03 secs
16384 1 mins 59 secs