Performance Impovement for Join

Question

I have the following join statement. I have 220,000 records ... when I run this program, it times out. Can any one suggest performance impovement?
  SELECT vbapvbeln  vbapposnr  vbapkwmeng  vbapmatnr
         vbapwerks  vbapkdmat  vbapposex   vbakauart
         vbakkunnr  vbakvdatu  vbepedatu   lipsvbeln
         lipsposnr  lipslfimg  marcdispo   marcbeskz "--> PV
         maktmaktx  marameins  vbkdihrez   vbkdempst
         vbkdbstkd  vbkdbstkd_e vbpa~ablad
  INTO TABLE lt_orders_t
  FROM vbak INNER JOIN vbap
              ON ( vbakvbeln = vbapvbeln )
            INNER JOIN vbep
              ON ( vbapposnr = vbepposnr AND vbapvbeln = vbepvbeln )
            LEFT OUTER JOIN lips
              ON ( vbapvbeln = lipsvgbel AND vbapposnr = lipsvgpos )
            LEFT OUTER JOIN marc
              ON ( vbapmatnr = marcmatnr AND vbapwerks = marcwerks )
            LEFT OUTER JOIN makt
              ON ( vbapmatnr = maktmatnr AND makt~spras = 'E'        )
            LEFT OUTER JOIN mara
              ON ( vbapmatnr = maramatnr )
            LEFT OUTER JOIN vbkd
              ON ( vbapvbeln = vbkdvbeln AND vbapposnr = vbkdposnr )
            LEFT OUTER JOIN vbpa
              ON ( vbapvbeln = vbpavbeln AND vbpa~posnr = '000010' AND
                   vbpa~parvw = 'WE'       )
  WHERE vbak~vbeln IN s_vbeln AND
        vbep~etenr = '0001'   AND
        vbap~posnr IN s_posnr AND
        vbap~werks IN s_werks AND
        vbak~auart IN s_auart AND
        vbak~kunnr IN s_kunnr AND
        vbak~vbtyp IN s_vbtyp AND
        vbap~matnr IN s_matnr AND
        vbep~ettyp IN s_ettyp AND
        vbep~edatu IN s_edatu.

Answer 1

Hi PV,
As others have mentioned you really need to break the statement down. 3 joins is usually pushing it (depending on the tables). By the way, you can go to SE12 and look up each table. On the main screen you will see a pushbutton saying Indexes I think. Click on it and it will show you the secondary indexes for that table. There may be some or none defined. They may be SAP defined or customer defined (Z).
Keep your inner joins on vbak, vbap and vbep. The resulting table will be used as your driver for the other tables.
You can create another inner join of mara, marc and makt.
Use your driver table as For All Entries for the other tables.
If you keep your driver table to be as shown, create another table with the first 5 fields.( vbeln, posnr, kwmeng, matnr, werks). The reason for this is so you can load the table with:
It_newtab[] = It_orders_t[].
All you want from the new table is matnr werks.
Sort it_newtab by matnr werks.
Delete adjacent duplicates from i_newtab comparing matnr werks.
Now you can use it_newtab as your driver for marc-mara-makt join. The reason is that if you have 220,000 records, when you narrow down the material/plant combinations you will probably have only a few hundred materials. So your For All entries will be a table of a few hundred instead of 200 thousand.
The result will be in a new table it_materials. When you read this table, don't forget to sort it by matnr and plant first. And READ with Binary Search.
Once you have all your other smaller tables created. As with it_materials, SORT and READ with BINARY SEARCH, as you loop through your driver table and populate the missing data from the smaller tables.
I think you will end up with 5 Select statements.
Indexes are important. Once you have split out the Select statements, if the code is still running unacceptably slow, you can use ST05 to see what indexes are being used and see if new indexes need to be created to help your cause.
You'll probably have to involve the Basis team to help you with your new indexes and they can help with the analysis of your ST05 results as well. Indexes must be created correctly in the proper order, or they may not be effective.
Hope this helps.
Filler