|
View previous topic :: View next topic
|
| Author |
Message |
haohao
New User
Joined: 23 May 2008
Posts: 35
Location: beijing China
|
|
|
|
hello,everybody here!
I have a performance issue about a batch program. when the BATCH program
is running, about 80% of the running time is waiting. from IBM APA report as bellow:
Overall CPU Activity
Samples 60,000 100.0% ' ' ' ' ' ' ' ' ' ' '
CPU Active 6,423 10.7% ll
WAIT 51,027 85.0% lllllllllllllllll
Queued 2,550 4.2% l
then,I CHECK the wait time report, as bellow:
| Code: |
W01: WAIT Time by Task/Category (0745/ESY100B2) Row 00001 of 00021
Name Description Percent of Time in WAIT * 10.00% 0.4%
*....1....2....3....4....5....6....7....8.
IKJEFT01-001 TCB=008DC7D8 100.00 lllllllllllllllllllllllllllllllllllllllll
IKJEFT02-006 TCB=008DC3D8 100.00 lllllllllllllllllllllllllllllllllllllllll
IKJEFT09-007 TCB=008DC148 100.00 lllllllllllllllllllllllllllllllllllllllll
DSN-008 TCB=008ACD50 100.00 lllllllllllllllllllllllllllllllllllllllll
DSNECP10-009 TCB=008ACBB8 85.24 lllllllllllllllllllllllllllllllllllllllll
SYSTEM System/OS 49.04 lllllllllllllllllllllllll
Services
SVC SVC Routines 49.04 lllllllllllllllllllllllll
SVC008 LOAD 49.04 lllllllllllllllllllllllll
IEAVEWAT Task 49.04 lllllllllllllllllllllllll
management
DATAMG Data Mgmt 28.13 llllllllllllll
Processing
DB2SQL SQL Processing 7.93 llll
|
the report shows that the SVC008 occupied the 50% of the waiting time.
from the ibm redbook, SVC008 is seems to be the system module to load the subprogram
into real storage. BUT it's very strange that the most of subprogram are static call by the
main program,so the subprogram should be link-edit into one loadmodule with the main program.
why so much LOAD still exist here? how can I reduce the waiting time?
I have no idea about that, so could anybody give me some suggestion?
thanks in advance!![/code] |
|
| Back to top |
|
 |
Bill Dennis
Active Member
Joined: 17 Aug 2007
Posts: 562
Location: Iowa, USA
|
|
|
|
There are other modules being loaded on your behalf, for COBOL run-time and system subroutines.
Reducing the time might involve faster DASD, ensuring the libraries are in the Library Look Aside (LLA) list and other issues. |
|
| Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008
Posts: 8697
Location: Dubuque, Iowa, USA
|
|
|
|
I'm not running APA, but from the manual (my emphasis added):
| Quote: |
Detail line descriptions
Each line represents a System Object - an object to which WAIT time is attributed. These lines are arranged hierarchically. You can expand a line (using the "+" line command) to reveal a breakdown into subordinate objects. Each type of object shown in this report is described here: |
so your statement
| Quote: |
| the report shows that the SVC008 occupied the 50% of the waiting time. |
really should be that SVC008 and all subordinate tasks took 50% of the waiting time. This is a radically different statement from what you claim -- and task management, for example, is one of those subordinate tasks. I think you're jumping to conclusions with inadequate data -- you need to look at everything included in the SVC008 number to know for sure what's eating up the time. The actual LOAD from disk could be a small fraction of the total. |
|
| Back to top |
|
|
haohao
New User
Joined: 23 May 2008
Posts: 35
Location: beijing China
|
|
|
|
actually,the SVC008 NODE only has this IEAVEWAT subtask.
this is the fullly expand view from APA W01 report:
| Code: |
DSNECP10-009 TCB=008ACBB8 85.24 *******************************************
> SYSTEM System/OS 49.04 *************************
Services
> SVC SVC Routines 49.04 *************************
> SVC008 LOAD 49.04 *************************
> IEAVEWAT Task 49.04 *************************
management
> IGC001 Supervis 49.04 *************************
or
Control
> DATAMG Data Mgmt 28.13 **************
Processing
> GLCPARAM VSAM 28.13 **************
> POINT UT0145+FFE 28.13 **************
> IDA019L1 Virtual 28.13 **************
I/O (VIO)
and VSAM
> IDA019R3 Virtual 28.13 **************
I/O
|
|
|
| Back to top |
|
|
haohao
New User
Joined: 23 May 2008
Posts: 35
Location: beijing China
|
|
|
|
from another report from APA,shows that over 30% disk I/O on LE load
library CEE.SCEERUN,show as bellow:
| Code: |
D03: DASD Usage Time by Dataset (0743/ESY100B1) Row 00001 of 00007
Dataset Name>DDName Percent of Time * 10.00% 1.2%
*....1....2....3....4....5....6....7
CEE.SCEERUN 38.10 lllllllllllllllllll
STEPLIB 1P1RS2 38.10 lllllllllllllllllll
BANCS.FNETLVEK.GLCPARAM 18.75 lllllllll
BANCS.FNETBTCH.LOAD.DBIO 5.73 lll
BANCS.FNETBTCH.LOAD.SUBPROG 3.31 ll
BANCS.FNETBTCH.LOAD 3.08 ll
|
it is very strange that why so much I/O on LE library. this libry is
already added to LINKLIST:
| Code: |
********************************* TOP OF DATA *****************************
57 12:36:13.00 ISF031I CONSOLE BAAP127 ACTIVATED
57 12:36:13.00 -D PROG,LINKLIST
57 12:36:13.05 CSV470I 12.36.13 LNKLST DISPLAY 197
LNKLST SET LNKLST06 LNKAUTH=LNKLST
ENTRY APF VOLUME DSNAME
1 A 1P1RS1 SYS1.LINKLIB
2 A 1P1RS1 SYS1.MIGLIB
3 A 1P1RS1 SYS1.CSSLIB
4 A 1P1RS1 SYS1.SIEALNKE
5 A 1P1RS3 SYS1.SIEAMIGE
6 A CP1CT1 SYS1.RACF.LINKLIB.NORMAL
7 A CP1CT1 SYS1.PLEXP1.LINKLIB
8 A 1P1RS3 ASM.SASMMOD1
9 1P1RS2 CEE.SCEERUN2
10 A 1P1RS2 CEE.SCEERUN
11 A 1P1RS3 CBC.SCCNCMP
|
|
|
| Back to top |
|
|
Bill Dennis
Active Member
Joined: 17 Aug 2007
Posts: 562
Location: Iowa, USA
|
|
|
|
| Is the library also in your STEPLIB? |
|
| Back to top |
|
|
haohao
New User
Joined: 23 May 2008
Posts: 35
Location: beijing China
|
|
|
|
| sure,the CEE LIBRY is in my job's step libry. what's the matter? |
|
| Back to top |
|
|
haohao
New User
Joined: 23 May 2008
Posts: 35
Location: beijing China
|
|
|
|
hi ,Bill:
I remove the CEE libry from the steplib ,the I/O on CEE is reduced almost to zero,now the batch program's preformance is imporved about 50% .
but would you please explain why ?
what I must pay attention to when I code the batch JCL steplib?
is it always better to remove all libraries in JOB STEPLIB who are already existed in LINKLIST?
thank you very much!! |
|
| Back to top |
|
|
|
|
