"matthieu exbrayat" <matthieu.exbrayat@lifo.univ-orleans.fr> wrote in
message news:bu0ofj$fkl$1@isis.univ-orleans.fr...
> Hi all,
>
> I am desperately looking for some technical information concerning the way
DB2 UDB is integrating
> parallelism.
>
> - which data distribution strategies are available?
> - how much is data distribution automatized ?
> - can the dba tune this?
> - etc.
>
> I would greatly appreciate if you could send me some docs or links...
>
> Thanks!
>
> Matthieu EXBRAYAT
> Associate Professor
> Univ. Orleans, France
>
There are several different ways DB2 effects parallelism:



1. Multiple database partitions (on the same node or multiple nodes). This
is the classic method of parallelism which is enables by distributing a
single table across multiple partitions, each working in parallel. This
requires DB2 ESE with Data Partitioning Feature.



2. Intra-partition parallelism - Parallelism within a single database
partition. This works well when a tablespace is stripped across multiple
physical disk drives.



3. Utility parallelism - This includes I/O parallelism for backups,
restores, creating indexes, etc.



4. Query parallelism enabled with Union All views. See this link for more
information:

http://www-106.ibm.com/developerwork...202zuzarte.pdf



5. In addition, the database manager itself can create multiple tasks to
perform the same function such as I/O cleaners (writing updated pages from
buffer pools to disk), and many others.



There is more information in the DB2 manuals. Here is the link to the
manuals:

http://www-3.ibm.com/cgi-bin/db2www/...bs.d2w/en_main

You can go to the "Master Index" manual and search on parallelism. The 3
Administration Guides (Planning, Implementation, and Performance) have good
general discussions of this topic, especially with regard to multiple
partitions.



On page 9 of the "Administration Guide: Implementation" manual, it describes
ways to enable parallelism in DB2.