Page 1 of 1

Inetrpreting performance stats

Posted: Wed Nov 15, 2006 1:04 pm
by asitagrawal
Hi plz help me interpreting this stats.
The Inter Process buffering was enabled. It was run on a single processor system.

Code: Select all

Trans_FSCM_2_EPM_LN:				
Name	Percent	Count	Minimum	Average
Collect_And_Sort.SortedData_To_Trans	92	5000	1	3174
R_JrnlSync_Hash.Max_JrnlSync_lkp	0	5000	2	3
R_MaxJrnlLines_Hash.Read_PreviousJrnlLine_Hash	0	5000	1	3
IPC_TGTJrnlLine.Trans_IPC_LoadJrnlLine	0	346	3	4
IPC_TGTJrnlLineSis.Trans_IPC_LoadJrnlLineSis	0	346	1	1
MaxJrnlLines_Hash.Update_JrnlLine_Hash	0	346	10	13
Status.Update	0	9	185	1142
Max_JrnlSync_lkp.Derivation	0	5000	10	13
Read_PreviousJrnlLine_Hash.Derivation	0	5000	9	12
StageVars.Derivation	0	5000	2	4
Trans_IPC_LoadJrnlLine.Derivation	0	346	82	99
Trans_IPC_LoadJrnlLineSis.Derivation	0	346	26	32
Update_JrnlLine_Hash.Derivation	0	346	9	10
Thanks

Posted: Wed Nov 15, 2006 1:51 pm
by ArndW
Where are these statistics coming from? Without a frame of reference it is impossible to do anything with these stats. What do you want to get out of them?

Posted: Wed Nov 15, 2006 1:58 pm
by asitagrawal
ArndW wrote:Where are these statistics coming from? Without a frame of reference it is impossible to do anything with these stats. What do you want to get out of them? ...
This is the perf statistics for a Transformer Stage.

Posted: Wed Nov 15, 2006 2:29 pm
by ray.wurlod
This is a tab-separated report obtained by enabling stage tracing from the Job Run Options dialog when issuing a job run request and checking the Statistics check box.

Copy it from there and paste it into a tool like Excel.

You can use it to identify the "hot spots" within stage execution.

Your statistics have been taken from a Transfomer stage called Trans_FSCM_2_EPM_LN. Each resource (link processing, derivations, evaluations of stage variables, updating stage status in the repository, etc.) shows in the report. You are (except for Count) looking at time; Minimum and Average times are given in rounded microseconds if my memory serves me correctly.

So your Transformer stage spent 92% of its time handling the Collect_And_Sort.SortedData_To_Trans link. You should probably be looking at tuning that part of the operation first. It may be masking problems downstream of itself, but you can't determine that till you've cleared that particular logjam.

Posted: Wed Nov 15, 2006 2:50 pm
by asitagrawal
ray.wurlod wrote:This is a tab-separated report obtained by enabling stage tracing from the Job Run Options dialog when issuing a job run request and checking the Statistics check box.

Copy it from there and paste ...
Well Ray, this is been copy and psted from the logs itself. I am sorry I can't allign them here..

Code: Select all

Name	Percent	Count	Minimum	Average
J..Collect_And_Sort.SortedData_To_Trans	0	2	9	22895
J..R_JrnlSync_Hash.Max_JrnlSync_lkp	0	2	6	8
J..R_MaxJrnlLines_Hash.Read_PreviousJrnlLine_Hash	0	2	4	4
J..IPC_TGTJrnlLine.Trans_IPC_LoadJrnlLine	0	2	8	10
J..IPC_TGTJrnlLineSis.Trans_IPC_LoadJrnlLineSis	0	2	3	3
J..MaxJrnlLines_Hash.Update_JrnlLine_Hash	0	2	11	38
Status.Update	99	2	6224	8410
Max_JrnlSync_lkp.Derivation	13	2	17	1428
Read_PreviousJrnlLine_Hash.Derivation	0	2	11	16
StageVars.Derivation	0	2	4	5
Trans_IPC_LoadJrnlLine.Derivation	2	2	94	158
Trans_IPC_LoadJrnlLineSis.Derivation	2	2	29	37
Update_JrnlLine_Hash.Derivation	0	2	10	11

Posted: Wed Nov 15, 2006 3:40 pm
by ray.wurlod
I wasn't asking you to; I was suggesting how you can make the data easier to view for yourself.