Performance: Open SQL vs. Native SQL (Oracle)

Question

Hi everybody,
I have an interesting issue here. For a DB selection I use an Open SQL query from a table view into an internal table. It works fine, but the performance is not very well. The SELECT uses LIKE and wildcards (%) to search for customer master data (names and address fields).
Because of the bad performance I made some tests in transaction DB02 with native SQL, but exactly the same SELECT structure. It looks like this:
  SELECT *
    FROM zzrm_cust_s_hlp
   WHERE client = 100
     AND mc_name1      LIKE '<name>%'
     AND mc_name2      LIKE '<name>%'
     AND valid_from    <= <timestamp>
     AND valid_to      >= <timestamp>
Ok, now I tried exactly the same SELECT statement with the same data to search for (<name> and <timestamp) with Open SQL and Native SQL. The Difference is quite suprising, the Native SQL query is about 5-10 times faster (arount 1 sec) than the Open SQL query (around 5-10 sec). Even with the LIKE keywords and the wildcards.
Any ideas what could be the problem with the Open SQL query?
And: what can I do to achive the same performance as with the Native SQL query?
Kind regards and thanks in advance for any help,
Matthias

Answer 1

Ok, here is the the SQL explaination from the DB02 query:
SELECT STATEMENT ( Estimated Costs = 194 , Estimated #Rows = 1 )
       9 COUNT STOPKEY
         Filter Predicates
           8 NESTED LOOPS
             ( Estim. Costs = 193 , Estim. #Rows = 1 )
             Estim. CPU-Costs = 1,665,938 Estim. IO-Costs = 193
               5 NESTED LOOPS
                 ( Estim. Costs = 144 , Estim. #Rows = 98 )
                 Estim. CPU-Costs = 1,162,148 Estim. IO-Costs = 144
                   2 TABLE ACCESS BY INDEX ROWID BUT000
                     ( Estim. Costs = 51 , Estim. #Rows = 93 )
                     Estim. CPU-Costs = 468,764 Estim. IO-Costs = 51
                     Filter Predicates
                       1 INDEX SKIP SCAN BUT000~NAM
                         ( Estim. Costs = 6 , Estim. #Rows = 93 )
                         Search Columns: 1
                         Estim. CPU-Costs = 59,542 Estim. IO-Costs = 6
                         Access Predicates Filter Predicates
                   4 TABLE ACCESS BY INDEX ROWID BUT020
                     ( Estim. Costs = 1 , Estim. #Rows = 1 )
                     Estim. CPU-Costs = 7,456 Estim. IO-Costs = 1
                     Filter Predicates
                       3 INDEX RANGE SCAN BUT020~0
                         ( Estim. Costs = 1 , Estim. #Rows = 1 )
                         Search Columns: 2
                         Estim. CPU-Costs = 3,661 Estim. IO-Costs = 1
                         Access Predicates
               7 TABLE ACCESS BY INDEX ROWID ADRC
                 ( Estim. Costs = 1 , Estim. #Rows = 1 )
                 Estim. CPU-Costs = 5,141 Estim. IO-Costs = 1
                 Filter Predicates
                   6 INDEX UNIQUE SCAN ADRC~0
                     Search Columns: 4
                     Estim. CPU-Costs = 525 Estim. IO-Costs = 0
                     Access Predicates
And this is the one from the Open SQL query in ABAP:
SELECT STATEMENT ( Estimated Costs = 15,711 , Estimated #Rows = 29 )
       7 NESTED LOOPS
         ( Estim. Costs = 15,710 , Estim. #Rows = 29 )
         Estim. CPU-Costs = 3,021,708,117 Estim. IO-Costs = 15,482
           4 NESTED LOOPS
             ( Estim. Costs = 15,411 , Estim. #Rows = 598 )
             Estim. CPU-Costs = 3,018,711,707 Estim. IO-Costs = 15,183
               1 TABLE ACCESS FULL BUT020
                 ( Estim. Costs = 9,431 , Estim. #Rows = 11,951 )
                 Estim. CPU-Costs = 2,959,067,612 Estim. IO-Costs = 9,207
                 Filter Predicates
               3 TABLE ACCESS BY INDEX ROWID ADRC
                 ( Estim. Costs = 1 , Estim. #Rows = 1 )
                 Estim. CPU-Costs = 4,991 Estim. IO-Costs = 1
                 Filter Predicates
                   2 INDEX UNIQUE SCAN ADRC~0
                     Search Columns: 4
                     Estim. CPU-Costs = 525 Estim. IO-Costs = 0
                     Access Predicates
           6 TABLE ACCESS BY INDEX ROWID BUT000
             ( Estim. Costs = 1 , Estim. #Rows = 1 )
             Estim. CPU-Costs = 5,011 Estim. IO-Costs = 1
             Filter Predicates
               5 INDEX UNIQUE SCAN BUT000~0
                 Search Columns: 2
                 Estim. CPU-Costs = 525 Estim. IO-Costs = 0
                 Access Predicates
Of course I can see the difference.
But since the statements are identical, I don't understand why this difference exists
Thanks for your help!
Kind regards, Matthias