MVSFORUMS.com

tattva · Beginner Joined: 02 Feb 2005 Posts: 97 Topics: 36

Hi,

I have changed a Cob program by removing the ORDER by clause in the SQL query.

Below are observations:-

Run time CPU time
Baseline 203.8 22.36
Changed 96.48 21.30
Improvement 52.65947007 0.268336315

Now as you see it... there is not much change in the CPU time taken. But there is significant
reduction in the Cycle time.

Now my question :-

Can i assume that the changed code is more efficient than the baseline code ( based on the
reduction in the cycle time)?

Note :- ORDER by clause was based on the INDEX only.
I Have replaced the oder by clause with SYCSORT step.

Snippet of the JESMSGLG:-

Baseline:-PROCSTEP RC EXCP CPU SRB CLOCK SERV
STEP0010 00 48340 22.36 .01 203.80 14478K

Changed :- PROCSTEP RC EXCP CPU SRB CLOCK SERV
STEP0010 00 48291 21.14 .01 83.17 13692K

Thanks,
Tattva

semigeezer · Posted: Fri Nov 18, 2005 3:28 pm Post subject:

That depends. Did you include the Syncsort step in these calculations? If not, then you are simply saying that it takes less time to do less work.

tattva · Beginner Joined: 02 Feb 2005 Posts: 97 Topics: 36

Yup..

I did take SYNCSORT step into the consideration.

Thanks,
Tattva

Phantom · Posted: Sat Nov 19, 2005 3:38 am Post subject:

Tattva,

External sort (whatever the product be) always is faster than DB2 Order By. Using this technique always depends on the way the cobol program is written. If it does lots of processing before the fetch and after the fetch-order by then you need to see if the program can be divided into two. Also, you may have to consider any future changes which may force you to use DB2 Order By.

If everthing seems fine, then go ahead with your new logic.

Thanks,
Phantom