****************************************************************

* TECH NOTE CD602.T19                               August 1996

*

* KEYWORDS: DB-EZReorg OPTIMISING SORTS

*

* DB-90S RELEASE 4.2 TAPES : C60201

*

* DB-EZReorg Sorts

*

* The sample JCL for DBEZRFWD provides 6 sort work files

* SORTWK01-06. For large sorts, ie if processing a number

* of journal files in one cycle which contain a high level of

* update activity to the affected areas, then this may not

* be most effective.

*

* To optimise the sorts:

* 1) Specify SORTWK01-03 only and size each at

*    approximately 60% of the size of the data being sorted

* 2) Allocate each to a seperate volume

* 3) Do not specify secondary extents

* 4) Specify REGION=0M on the EXEC card

* 5) Ensure the input and output datasets are on different volumes

*    to each other and the sort work files.

*

* For example, if 600 cylinders of data are to be sorted specify

* //SORTWK01 DD UNIT=SYSDA,SPACE=(CYL,360)

* //SORTWK02 DD UNIT=SYSDA,SPACE=(CYL,360),SEP=SORTWK01

* //SORTWK03 DD UNIT=SYSDA,SPACE=(CYL,360),SEP=(SORTWK01,SORTWK02)

*

* Sizing the sort work files can only be guessed before the run

* as it depends on the volume of data processed. However if you

* specify STOPAFTER=DBEZCUP1 at the start of each cycle and then

* obtain the space used by the file DBEZJNLU this will tell you how

* much data will be sorted by the internal step DBEZSJNL. As the DML

* file generated by DBEZCUP2 will be around 30% of the size of

* DBEZJNLU you can also estimate the space required for this file

* and hence DBEZJNLU.SORT.

*

* If the output from IDMSDBL2 contains a large number of database

* keys (check the space used by DBEZDBK2) then the sort work files

* for DBEZSXT3 and DBEZSXT4 may also need to be increased. You

* should also ensure that the blocksize of DBEZDBK2, DBEZDBK3

* and DBEZDBK4 is close to half track.

*

* If you are suffering DASD constraints then you may minimise

* the DASD requirements through use of STOPAFTER. Eg the file

* DBEZJNLU may be reallocated as 1 track after DBEZJNLS as the

* sorted data is now in DBEZJNLU.SORT. Likewise DBEZDML may be

* allocated initially as 1 track, then properly sized after

* DBEZJNLS and reallocated back to 1 track after DBEZDMLS.

*

****************************************************************