Performance problem on query

Question

As part of a large batch process, we load data sent from a client into our systems. The data consists of information representing their organization structure (departments, users etc.). I need to identify all users that are currently in the existing system that are not part of the new file. The second step of the process just won't perform, so I"m looking for suggestions. Here is the psuedocode and then the SQL:<br/><br/>
* 1. Load all existing users for the client into a <i>global temporary table</i>.<br/>
* 2. Delete all users from the temp table that are not on the file.<br/>
* 3. Delete all top level users from the temp table (they can't be deleted).<br/>
* 4. Delete all users from the table that already have the deleted role.<br/>
* 5. Any remaining rows are the users that need to be deleted.<br/><br/>
DDL to create table<br/>
--------------------------------<br/><br/>
CREATE GLOBAL TEMPORARY TABLE DELETE_USERS_TEMP (
EMPLOYEE_ID VARCHAR2 (100),
USER_ID VARCHAR2 (100)
) ON COMMIT DELETE ROWS;<br/><br/>
1. insert into DELETE_USERS_TEMP select employee_id, user_id from cadreadmin.orderentry_usr_data where user_id like ? and employee_id is not null;<br/>
2, delete from DELETE_USERS_TEMP where employee_id in (select emplid from client_dataset_detail where cd_id = ?);<br/>
3. delete from DELETE_USERS_TEMP where user_id in (select USER_ID from CADREADMIN.DPS_USER_ORG where USER_ID like ? and ORGANIZATION = ?);<br/>
4. delete from DELETE_USERS_TEMP where user_id in (select user_id from cadreadmin.dps_user_roles where atg_role = ? and user_id in (select user_id from DELETE_USERS_TEMP));<br/><br/>
I have also tried the following for step 2:
delete from DELETE_USERS_TEMP T
where exists (select D.EMPLID from CLIENT_DATASET_DETAIL D
where D.CD_ID = 170 and D.EMPLID = T.EMPLOYEE_ID);<br/><br/>
Here is the explain plan for the original step 2 SQL, the one just above is about the same:<br/><br/>
"Optimizer"     "Cost"     "Cardinality"     "Bytes"     "Partition Start"     "Partition Stop"     "Partition Id"     "ACCESS PREDICATES"     "FILTER PREDICATES"
"UPDATE STATEMENT"     "ALL_ROWS"     "7"     "1"     "76"<br/>     ""     ""     ""     ""     ""
"UPDATE FILEUPLOAD.DELETE_USERS_TEMP"<br/>     ""     ""     ""     ""     ""     ""     ""     ""     ""
"NESTED LOOPS(SEMI)"     ""     "7"     "1"     "76"<br/>     ""     ""     ""     ""     ""
"TABLE ACCESS(FULL) FILEUPLOAD.DELETE_USERS_TEMP"     "ANALYZED"     "2"     "1"     "65"<br/>     ""     ""     ""     ""     ""
"TABLE ACCESS(BY INDEX ROWID) FILEUPLOAD.CLIENT_DATASET_DETAIL"     "ANALYZED"     "5"     "10770"     "118470"<br/>     ""     ""     ""     ""     ""D"."CD_ID"=170"
"INDEX(RANGE SCAN) FILEUPLOAD.CLIENT_DATASET_DETAIL_INX2"     "ANALYZED"     "2"     "3"<br/>     ""     ""     ""     ""     ""D"."EMPLID"="T"."EMPLOYEE_ID""     ""
Message was edited by:
user620465

Answer 1

Pls, take a look at When your query takes too long ... . <br>
Also, in step 1 you can avoid the copy of top level users and the ones that already have the deleted role. With a smaller table, the response time of step 2 can be reduced.
Regards,
Miguel