Sunday, June 17, 2012

Query Performance

Do you know columns in selection list affect use of index on table..??

I couldn't  believe this..I was always under impression that query plan is decided based on from and where clause and select clause has no impact on index usage. I was proved wrong in the Smoke Test for INT2 at midnight yesterday..

See how...

select trx_number, org_id, interface_header_attribute1      
          from ra_customer_trx_all
          where org_id = 112 
         
Plan
SELECT STATEMENT  ALL_ROWSCost: 1,947  Bytes: 219,538  Cardinality: 9,979     
    1 TABLE ACCESS FULL TABLE AR.RA_CUSTOMER_TRX_ALL Cost: 1,947  Bytes: 219,538  Cardinality: 9,979 



select trx_number, org_id      
          from ra_customer_trx_all
          where org_id = 112
         
Plan
SELECT STATEMENT  ALL_ROWSCost: 551  Bytes: 129,727  Cardinality: 9,979     
    1 INDEX SKIP SCAN INDEX AR.RA_CUSTOMER_TRX_N1 Cost: 551  Bytes: 129,727  Cardinality: 9,979 
        
Ultimately the solution to this performance issue was add one more table to query... Interesting huh?

bad query
select trx_number
          from ra_customer_trx_all
          where 1=1
          and interface_header_attribute1 = to_char(pin_order_number)
          and org_id = v_Header_Rec_Type.org_id
          and rownum = 1;

Good query

select trx_number      
          from apps.ra_customer_trx_all rct, apps.ra_customer_trx_lines_all rctl
          where rct.customer_trx_id = rctl.customer_trX_id
            and rctl.sales_order = to_char(66000120)
            and rct.org_id = 112
            and rownum < 2;

No comments:

Post a Comment