Help tuning the query on NAST table .

Question

I have the below query on NAST table which is performin very badly:
SELECT objky INTO TABLE i_objky
     FROM nast
     WHERE kappl = c_kappl            
     AND   ( ( kschl = c_kschl_zbax ) 
           OR ( kschl = c_kschl_zoem ) )
     AND   vstat = c_vstat_comp       
     AND   objtype = c_likp
      AND ( ( datvr EQ w_start_date AND
              uhrvr GE w_start_time ) OR
            ( datvr EQ w_end_date AND
              uhrvr LE w_end_time ) OR
            ( datvr GT w_start_date AND
              datvr LT w_end_date ) ).
Further there is already an index on KAPPL and KSCHL.
Message was edited by:
        Ashish kumar

Answer 1

If I understood your logic correctly - you are trying to select objects which have outputs processed within your date/time interval.
You can try to rewrite your select to force index 'A' usage as suggested by Christian, but it is not very helpful if you have too many records in NAST as in this case this index is not very selective and optimizer usually ignores it. Of course, you can ask your BASIS folks to tell you if you can use 'hint' addition in your select to force index usage as it depends on DB and some DBs require some work to be done on DB level to enable hints.
Other options could be
- try to provide a range (even if it is wide enough) for object key... it may really help as NAST contains too many records for different objects and this range may help with primary key usage.
- OR if you are interested in these 2 output types only - you can increase the speed of your select significally if you define a separate database table which you will update with these outputs (when they are just created in NAST) and your program (if it usually runs just once per your date.time range) will check the status of these outputs and then delete records from your table or just update them with "processed" status.
You can actually just create a copy of NAST table which is going to be updated when these outputs are processed (from corresponding processing program), so you will have records there with at least date, and then based on these records you can select actual date/time from NAST. It will be much faster... but it requires more design/coding.