OLAP Query is Taking Time. Please give me the Solution.

Question

Hi Geof,
I am new to this AW Cubes and all. I am facing with a problem. After reading u r responses and all I am asking u this question. I hope u will solve my problem.
I trying to get the data out of cube using the below statement at SQL prompt.
SELECT CONSTITUTION,ACTIVE_CODE,CLASSIFICATION,BR_CODE,CASTE,INTEREST_RATE,TOT_COUNT,OUTSTANDING_BAL FROM TABLE(OLAP_TABLE('MIS.KBL_REPORTS_AW DURATION SESSION',
'DIMENSION CONSTITUTION AS varchar2(100) FROM AW_CUSTOMER_DIM1
DIMENSION ACTIVE_CODE AS varchar2(50) FROM AW_ACTIVITY_DIM1
DIMENSION CLASSIFICATION AS varchar2(50) FROM AW_PRODUCT_DIM1
DIMENSION BR_CODE AS varchar2(50) FROM AW_BRANCH_DIM1
DIMENSION CASTE AS varchar2(50) FROM AW_CASTE_DIM1
DIMENSION INTEREST_RATE AS varchar2(50) FROM AW_INTEREST_DIM1
MEASURE TOT_COUNT AS number(5) FROM KBL_REPORTS_CUBE1_COUNT_MEASURE
MEASURE OUTSTANDING_BAL AS number(10) FROM KBL_REPORTS_CUBE1_BALANCE_MEASURE
where CONSTITUTION like '%INDIVIDUAL%' and ACTIVE_CODE like '%010101';
It is taking too much of time. Can u give the solution to me ???
Regards,
S.Vamsi Krishna

Answer 1

Hi,
What release are you using? You should use 10.2.0.2 if possible. The documentation references in this reply assume the 10.2.0.2 versions of the documentation.
You may be able to add an empty MODEL clause to your query which has the paradoxical effect of speeding up most OLAP queries. See chapter 11 (OLAP_TABLE) in the Oracle OLAP Reference. But if performance is very poor, this is unlikely to improve matters enough.
We need to figure out where the time is being spent. We can use Oracle trace files and the tkprof utility to do that. I'll describe the basics, but there is a lot more information in chapter 20 (Using Application Tracing Tools) of the Oracle Database Performance Tuning Guide.
First, let's enable some configuration parameters that will cause the RDBMS to write some useful information to the trace file. Assuming that you have a SQL script that executes the query, you can add these commands to the beginning of the script:
alter session set sql_trace=true
alter session set timed_statistics=true
alter session set "_olap_continuous_trace_file"=true
alter session set "_olap_table_function_statistics"=true
alter session set tracefile_identifier=perftest
exec dbms_aw.execute('dotf tracefile');
Now run the script with the query.
Next we have to find the trace file. Trace files are written to the directory named by the configuration parameter user_dump_dest. As a privileged user, execute this command:
SQL> show parameter user_dump_dest;
Now look in that directory for a file whose name includes the string that we specified in the tracefile_identifier setting above (you can change this if you want):
$ ls <user_dump_dest dir>/*perftest*.trc
Having located the tracefile, run the tkprof utility:
$ tkprof <trace file name> <tkprof output file name> aggregate=no
Now edit the generated <tkprof output file name> and look for your query. It will be followed by a mini explain plan that includes a line for each subquery. You'll be able to see how much time is spent in each subquery and how many rows are returned to the enveloping query. The innermost subqueries will be the ones that query AW data. Your goal is to push as much of the work into the AW as possible, so that the AW engine return a minimal number of rows.
You can get more detail from the trace file itself. If you are at all familiar with the OLAP DML, look for LIMIT commands and see that they roughly correspond to hierarchy selections and WHERE clause predicates in the SQL SELECT statement. OLAP DML comamnds have timestamps in the trace file, so you can get an idea of how long things are taking. You can find table function statistics by looking for XSTF and XSTABFUN in the trace file.
Geof