| Re: Slow Sql that access Large Data [message #372593] |
Mon, 26 February 2001 14:34 |
Joachim Lindner
Messages: 30
Registered: February 2001
|
Member |
|
|
From my experience in our current project I can confirm the benefits of using
Oracle's PQO (Parallel Query Option) for large queries, in general. We even managed
to reduce the runtime of a complex statement by 50% on a single processor machine by
making 4 query servers process the query. As to be expected the performance boost of
using PQO on a multi-processor system is much higher, I do not have precise figures
here, however. What you should not forget is to increment the value setting for parameter
PARALLEL_MAX_SERVERS if you want more than 5 (default) query servers to be used. You
also have to keep in mind that resource consumption (sort area size etc.) multiplies
by the degree of parallelism you request.
In addition, my suggestion for high-performance processing of huge table joins is
to rely on HASH JOINS. In our project a self-join on an 80 million row table performs
best using the USE_HASH hint, although there is a B*Tree-index on the column in the
join condition.
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
