Multithreading in Oracle

Question

Hi All,
I am working on Oracle 11g. I have to implement multithreading/asynchronous processing in one of my applications.
There is a table, tab1 which contains two columns - ID and SQL. The SQL column is of CLOB datatype and contains a Dynamic SQL statement. This table has around 100 records and can contain upto 1000 (not more than that). The intent of these Dynamic SQLs is to update one of the tables and the dynamic sql is quite big (creates temp tables, business logic application, data is merged and finally updates are done, again based on rules). As of now, the application was executing these queries in sequential manner, through a simple loop.
I want to explore how can I introduce multithreading here, so that I can run the sqls in multiple threads. I know, I can use DBMS_SCHEDULER, but I am struggling at - how to break those 100 records into say 10 sessions at a time? I can track the IDs which have been processed by updating an update_date column in the above mentioned table.
Any suggestion will be highly appreciated which can put me in right direction.
If the problem statement is not clear, do let me know.

Answer 1

Hi,
I have never used before DBMS_SCHEDULER but the system parameter job_queue_processes should define the maximum number of jobs running at the same time. If set to 0 Oracle will determine it.
I have prepared a sample dummy code that is creating 50 jobs and starting them.
The parameter job_queue_processes is set to 10.
From the log table mylog you will see only 10 jobs are executing at the same time.
-- Table to store my SQL jobs
CREATE TABLE myjobs
   id             NUMBER NOT NULL
, SQLCODE        CLOB
-- Table to store log of my jobs
CREATE TABLE mylog
   start_time     TIMESTAMP NOT NULL
, end_time       TIMESTAMP NOT NULL
, id             NUMBER NOT NULL
-- Creating entries into myjobs table
BEGIN
   FOR i IN 1 .. 50
   LOOP
      INSERT INTO myjobs
           VALUES (
                     i
DECLARE
   v_starttime  TIMESTAMP;
   v_endtime    TIMESTAMP;
BEGIN
   v_starttime := SYSTIMESTAMP;
   dbms_lock.sleep(10);
   v_endtime := SYSTIMESTAMP;  
   INSERT INTO mylog VALUES (v_starttime, v_endtime,' || i || ');
   COMMIT;
END;');
   END LOOP;
   COMMIT;
END;
-- This will set maximum number of jobs at the same time
ALTER SYSTEM SET job_queue_processes=10;
-- Submitting the jobs
BEGIN
   FOR c1 IN (  SELECT *
                  FROM myjobs
              ORDER BY id)
   LOOP
      -- Delete the job if it is already existing with the same name
      BEGIN
         DBMS_SCHEDULER.drop_job ('test_sched_job_' || c1.id);
      EXCEPTION
         WHEN OTHERS
         THEN
            NULL;
      END;
      DBMS_SCHEDULER.create_job ( job_name => 'test_sched_job_' || c1.id
                                , job_type => 'PLSQL_BLOCK'
                                , job_action => c1.SQLCODE, enabled => FALSE);
      DBMS_SCHEDULER.run_job (job_name => 'test_sched_job_' || c1.id, use_current_session => FALSE);
   END LOOP;
END;
-- After all jobs have been completed
-- this will show the start time of jobs
-- 10 jobs at the same time have started as per setting
  SELECT *
    FROM myjobs
ORDER BY starttime, id;
/* Cleanup */
DROP TABLE myjobs;
DROP TABLE mylog;
{code}
Regards.
Al