[Info-Ingres] CREATE INDEX parallel vs CREATE INDEX in sequence

[jpzuate at gmail.com]

Hi all,

I have a table with 6 382 316 tuples and 7 indexes, 2 K pages. Table
is Heap.
Ingres version is II 9.0.4 (a64.lnx/105)NPTL on RedHat.

To rebuild indexes, I try two method :
- "old school" : CREATE INDEX in sequence
- "new style" : CREATE INDEX with parallel syntax

Result :
- CREATE INDEX in sequence : 17 min 18 sec
- CREATE INDEX parallel : 1 h 6 min 57 sec

It is the first time I test this technique with a large amount of data
and previous tests (with max 1 000 000 tuples) did not show a so large
difference. Parallel syntax is supposed faster, better, and so on.

For information, partitionning the same table with 2 dimentions and 42
partitions take only 4 min 30 sec ...

Do we miss something in the configuration or something else ? Who have
experience to share ?

Thank in advance,
Jean-Pierre