One thing that helped was to compress the indexes (CAUTION: ON UNIQUE
INDEXES DO NOT COMPRESS = TO THE # OF COLUMNS, COMPRESS = TO THE # OF
COLUMNS -1). If the index was used then it was faster (usually) and
required less block scans. Compress the indexes AFTER you do an initial
data load.
Jim
"Andy" <enzoweb_at_hotmail.com> wrote in message
news:8d4033cd.0201142019.c7ff06f_at_posting.google.com...
> More info:
>
> optimizer_mode=rule (siebel won't let us use cost/choose; no
> point in analyzing tables
>
> db_block_buffers=97656
> hash_area_size = 50000000
> hash_join_enabled = TRUE
> log_buffer = 104857600
> optimizer_index_cost_adj=10
>
>
> Also, the Siebel guy said we can't use hints.
>
> Furthermore, some tables have over 20 indexes on them (the record is
> 33) and the Siebel guy says we shouldn't drop any.
Received on Mon Jan 14 2002 - 23:01:25 CST
