Timing Query Performance in Java

Question

Hello again everyone,
I had previously asked a question concerning timing the execution time of an individual thread, and I have been unable to uncover an acceptable answer. So, instead of threads I may use system processes. Some background...
I am writing a database load-testing application that simulates load via. submitting a set of queries through multiple connections. These multiple connections can be separate threads or processes that run concurrently in order to get a feel for how the database performs under the load of a number of users. My question is does anyone know how to get the actual execution time of a single process. I am afraid that due to the large number of users, the system will be unacceptably slow. Normally I could do something like this ...
long start = System.getcurrentTimeMillis();
executeJDBCquery();
end = System.getcurrentTimeMillis() - start;
however I am concerned that with a large number of users this will not provide an accurate query execution time since the system will be distrubting its processing power over so many processes. The first suggestion that has raised is to get the query processing time from the RDBMS, however, I can't figure out how this can be done with Oracle 8i. Is there a hidden JDBC method or property that will give server-side execution time for a query, or can anyone suggest a solution that will work with oracle? If anyone knows how to get the actual execution time of a single system process that would be very helpful as well. Thanks for any help that you may be able to give, let me know if I can provide any more details.
Dan

Answer 1

Try this:
Create a new table in Oracle to track the times, such as:
Create Table time_log (Thread_ID number, Query_ID number,
start_time number, end_time number);
On the tables that you are performing the queries on, create before and after statement-level triggers. These triggers would insert a record into the time_log field as required (i.e., the before trigger enters a start time, the after trigger enters an end time).
To properly keep track of the execution times, you would need a thread_id (possibly pass this as part of the query call) and a table/query_id. This way you would know which thread queried what table.
For your final timing computations, simply query the time_log table, making the correct calculations.
One note: the start and end times are numbers, not dates, as the Oracle dates only get down to seconds. You will have to use the DBMS_UTILITY.GET_TIME, which returns the current time in hundreds of a second.
HTH,
Sean