Hi All !

Can somebody explain the difference between the concept of partitioned tablespace in DB2 for z/OS (or any mainframe o/s) and DB2 for LUW (any non-mainframe o/s) ?
Or, tell me some links/documentation that best describes the partioned tablespace as implemented by DB2 for LUW. Do and Don'ts, DBA Maint specifics, Tips and Tricks etc for a partioned TS in DB2 for LUW.

Thanks
Anil

Hi Anil,

Hope bwlow like will helps you to understand partition TS,


http://publib.boulder.ibm.com/cgi-bi...STHIT#FIRSTHIT


Cheers,
Anguraj Rathinasamy.

Hi Anil,
shortly what I remember, not having used EEE (or DPF as it is called today):
on LUW you can have partitions over different nodes only (EEE) and they will always be hashed.
Range partitioning - even so having been announced for V8.2 or so - is not supported yet I think.

On z/OS a partitioned tablespace will hold a partitioned table, the range partitioning will be defined using the partitioning index. No hashing available yet, only range partitioning.

Sizing: on z/OS each partition can have a different size if needed - on LUW size should be (I think even has to be) the same due to hashing, but data schewing cannot always be avoided e.g. for default values (bad design but happens)

Please ask if it is too confusing.

As menitioned above, partitioned tablespaces in DB2 for z/OS is called range partitioning. This feature will be available in DB2 for LUW version 9.1 later this year (beta is available now). Range partitioning is primarily used for administrative ease of large tables, such as where one partition can be down for a load, while the other partitions can be on-line. You can also archive one partition at a time with a partition backup.

DB2 for LUW supports hash partitioning where each partition is on a seperate node (logical or physical) primarily for the purposes of parallel processing of a singe long-running SQL statement. The more different physical nodes there are, the more linear scalability one can achieve with the parallelism.

There are some exceptions to the above, such as the fact that parallel processing can be done on range partitioned tables, but the above is primary difference between the two.