External Loader Table - Merging Data

Question

Hi,
We have a legacy system that is Cobol based and it produces flat files that we ftp into a directory on our Oracle server.
I have created an external loader table, that has the location set to the specific files in that directory - currently holding about 5.6 million records.
The files contain patient data from mulitple sites, so I am trying to bring them all together into three different tables;
1) Patient_Id
2) Patient
3) Patient_Address
Patient_Id holds one global identifier for each patient (GUI).
I have written three merge statements, but the performance seems to be pretty shocking. An example of one is below:
    --MERGE RECORDS INTO E2E_PATIENT WHERE SITE ID MATCHES THAT IN FILENAME
    MERGE INTO E2e_Patient p
    USING (SELECT p.Pat_Id,
                  m.Title,
                  m.Forenames First_Name,
                  m.Surname,
                  m.Gender,
                  Pkg_Private_E2e.Fnc_Convert_Mdx_Dob(m.Dob,
                                                      'YYYYMMDD') Dob
           FROM   Mdx.e_Mpieghg1 m
           INNER  JOIN Commons.t_Site_Codes Sc
           ON     Sc.Medax_Site_Code = m.Site
           INNER  JOIN E2e_Patient_Id p
           ON     p.Local_Pas_Id = m.Mpi
           AND    p.Site_Id = Sc.Siteid
           AND    p.Sorce = 'MDX') m
    ON (p.Pat_Id = m.Pat_Id)
    WHEN MATCHED THEN
      UPDATE
      SET    p.Title      = m.Title,
             p.First_Name = m.First_Name,
             p.Surname    = m.Surname,
             p.Gender     = m.Gender,
             p.Dob        = m.Dob
    WHEN NOT MATCHED THEN
      INSERT
        (Pat_Id,
         Title,
         First_Name,
         Surname,
         Gender,
         Dob)
      VALUES
        (m.Pat_Id,
         m.Title,
         m.First_Name,
         m.Surname,
         m.Gender,
         m.Dob);
    COMMIT;If I run the merge on exactly the same data, it seems to be updating all records again - even though there are no changes - is this how the merge works?
Are there any other ways of increasing performance?
Regards
Mark
Edited by: Lloydy76 on 04-Feb-2011 04:07

Answer 1

Dan's approach might be a better way, but if there are a fairly large number of records coming in that do not need to be updated or inserted, then another approach might be to add a predicate to the when matched clause to avoid the updates when nothing has changed. Something like:
WHEN MATCHED THEN
   UPDATE
   SET p.Title = m.Title,
       p.First_Name = m.First_Name,
       p.Surname = m.Surname,
       p.Gender = m.Gender,
       p.Dob = m.Dob
   WHERE p.Title != m.Title or
         p.First_Name != m.First_Name or
         p.Surname != m.Surname or
         p.Gender != m.Gender or
         p.Dob != m.DobJohn