Delete overlapping requests runs for 5+ hours ...This is too long

Question

We are on BW 3.5 and have a process chain that loads 3 years of data from a cube on the APO Server and AOP data from the BW server into a cube on the BW server.  Each load is in its own infopackage.  This loads happens every week.  We want to delete the loads that occurred the week before using the delete overlapping requests.  It is set up as follows:
<b>1.</b>Drop index of Cube
<b>2</b>.Load Current Year data into Cube via Infopackage #1.  Loads approximately 32 million records.  Selections on calweek: 200701 – 200752.  Infosource: InfoSource for APO-DP Cube from APO(ZAPODP_I11). Datasource: APO-DP Backup Cube(8ZAPODPC73).  Source System: APP Client 200(APPCLNT200).  The Data targets tab is set up as follows: ‘Automatic Loading of similar/identical requests from the cube’ wherein Delete Existing Requests Conditions: Infosources are the same.  Selections are overlapping.
<b>3</b>.Load Current Year + 1 data into Cube via Infopackage #2.  Loads approximately 32 million records…Selections on calweek: 200801 – 200852.  Rest is same as above #2.
<b>4.</b>Load Current Year +2 data into Cube via Infopackage #3.  Loads approximately 32 million records…Selections on calweek: 200901 – 200952. .  Rest is same as above #2.
<b>5</b>.Load AOP data into Cube via Infopackage #4.  Loads approximately 135,000 records.  Infosource: AOP Plan(8ZAOPC01). Datasource: AOP Plan(8ZAOPC01). Source System: BW Production(BWPCLNT100).  The Data targets tab is set up as follows: ‘Automatic Loading of similar/identical requests from the cube’ wherein Delete Existing Requests Conditions: Infosources are the same.  Selections are overlapping
<b>6.</b>Create index on cube
<b>7</b>.Delete overlapping requests from cube.  The process chain step is set up to ‘Use Delete Selections for Infocubes from the Infopackages’
<b>The issue is that step #7 runs for more than 5 hours.</b>  That is way too long.  When we manually delete the requests, it takes 5 minutes.  How can I fine tune this processing?  Thanks in advance!
<b></b>

Answer 1

What is your dbms? Does it support partitioning? If it does then deleting requests merely drops the corresponding partitions, which should be very fast.
If it does not support partitioning then you're running SQL delete statements to delete the data, which could take a long time if you're deleting tens of million of records. BTW how long does it take to load the same data?
P.S. Just saw your last point about manually deleting the requests taking only 5 minutes. Did you delete one request at a time or multiple request at once? The previous poster wanted to know whether you had background processes to run the deletion job. i.e. Did the deletion really took 5 hours, or did it wait a few hours to start the job?
Message was edited by:
        Rick Chau