View previous topic :: View next topic
|
Author |
Message |
nickwyw
New User
Joined: 15 Jun 2008 Posts: 9 Location: China
|
|
|
|
Hi,
who can tell me, how to improve the system's performance, when we all logon the system on the same time, and someone compiler the programs, the system will be very slow.
How to find the limit of system performance bottleneck?
/D ASM
RESPONSE=TST
IEE200I 15.14.26 DISPLAY ASM 639
TYPE FULL STAT DATASET NAME
PLPA 93% OK PAGE.TST.PLPA
COMMON 6% OK PAGE.TST.COMMON
LOCAL 42% OK PAGE.TST.LOCAL1
LOCAL NV 37% OK PAGE.TST.LOCAL2
LOCAL 97% OK PAGE.TST.LOCAL3
LOCAL 23% OK PAGE.TST.LOCAL5
LOCAL 24% OK PAGE.TST.LOCAL4
By the way, how to show the number of memory used by a program? |
|
Back to top |
|
|
expat
Global Moderator
Joined: 14 Mar 2007 Posts: 8796 Location: Welsh Wales
|
|
|
|
Performance and capacity analysis is not a quick and easy job. You will need to analyse the SMF records to get a clear picture of what happened exactly when to cause your problems.
Then you can make changes, system settings, dataset movement etc. etc. qnd then start the monitor / analyse process all over again. |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19243 Location: Inside the Matrix
|
|
|
|
Hello,
Quote: |
how to improve the system's performance, when we all logon the system on the same time |
An analogy. . .If your system can carry a 100 unit load smoothly, but "when we all logon" a 750 unit load is created, everything will be slow. . .
Has the system always been slow when everyone logs on or is this a new behavior?
Have you added to the number of people logging on (especially developers) recently?
Excess capacity is a good thing to use when available, but adding more usage to an already full system is quite counter-productive. |
|
Back to top |
|
|
nickwyw
New User
Joined: 15 Jun 2008 Posts: 9 Location: China
|
|
|
|
Thanks,
yes, this is our traning system, we have about 80 new members practise developer program on it, I think the system is overload. Maybe we should upgrade the hardware. |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19243 Location: Inside the Matrix
|
|
|
|
Hello,
Quote: |
Maybe we should upgrade the hardware. |
Possibly, or change some of the other/scheduled work on the system to be done when the developers are not active.
80 new developers could surely cause a performance "spike". . . |
|
Back to top |
|
|
Bill Dennis
Active Member
Joined: 17 Aug 2007 Posts: 562 Location: Iowa, USA
|
|
|
|
nickwyw wrote: |
By the way, how to show the number of memory used by a program? |
Look for the IEF374I msg in your JCL listing. VIRT is memory below 16MB and EXT in memory above 16MB. |
|
Back to top |
|
|
Craq Giegerich
Senior Member
Joined: 19 May 2007 Posts: 1512 Location: Virginia, USA
|
|
|
|
Are the developers doing compiles in foreground or background? |
|
Back to top |
|
|
nickwyw
New User
Joined: 15 Jun 2008 Posts: 9 Location: China
|
|
|
|
our system have 512M memory, and OS/390, I'm not sure whether it support 80 users.
Quote: |
Look for the IEF374I msg in your JCL listing. VIRT is memory below 16MB and EXT in memory above 16MB. |
I find the COMPILER JCL LOG, the msg as follow:
IEF374I STEP/LKED /STOP 2008198.0914 CPU 0MIN 00.07SEC SRB 0MIN 00.01SEC VIRT 172K SYS 280K EXT 1316K SYS 10056K
This configuration reasonable?
Quote: |
Are the developers doing compiles in foreground or background? |
we compiler the program by submit a JCL. |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19243 Location: Inside the Matrix
|
|
|
|
Hello,
Quote: |
I find the COMPILER JCL LOG, the msg as follow:
IEF374I STEP/LKED /STOP |
This is from a compile/link. You would also want to find this info for batch program executions to see how much those jobs used.
Once programs are compiled, how are they run? CICS, IMS, batch, etc? Does this system also run db2? |
|
Back to top |
|
|
nickwyw
New User
Joined: 15 Jun 2008 Posts: 9 Location: China
|
|
|
|
This system also run DB2 and CICS.
have batch programs and online program. they read data from db 2 or data sets. online program we run them on the CICS, and submit a JCL to run the batch program.
follow are msg in compiler job log.
IEF374I STEP/PC /STOP 2008198.1020 CPU 0MIN 00.08SEC SRB 0MIN 00.01SEC VIRT 1320K SYS 296K EXT 4K SYS 10164K
IEF374I STEP/TRN /STOP 2008198.1020 CPU 0MIN 00.10SEC SRB 0MIN 00.00SEC VIRT 8332K SYS 288K EXT 4K SYS 9920K
IEF374I STEP/COBOL /STOP 2008198.1020 CPU 0MIN 00.34SEC SRB 0MIN 00.02SEC VIRT 9928K SYS 300K EXT 1845424K SYS 10076K (this is in a online program compiler)
IEF374I STEP/COBOL /STOP 2008198.1014 CPU 0MIN 00.16SEC SRB 0MIN 00.01SEC VIRT 8192K SYS 284K EXT 32720K SYS 10736K
(this is in a batch program compiler)
IEF374I STEP/LKED /STOP 2008198.1014 CPU 0MIN 00.07SEC SRB 0MIN 00.01SEC VIRT 172K SYS 280K EXT 1316K SYS 10056K |
|
Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008 Posts: 8700 Location: Dubuque, Iowa, USA
|
|
|
|
Hardware upgrade would definitely help. z10 boxes can go to 1.5 terabytes of memory; even our z800 is running 3 gigabytes per LPAR. 512 megabytes isn't that much these days, especially when running a lot of TSO sessions -- I do a lot of Unix Systems Services support so my TSO session memory limit is set to 128 MB. Additional memory would cut the paging shown in your initial post, which would also improve response times and performance. Upgrading OS/390 to z/OS would probably improve things somewhat, too. |
|
Back to top |
|
|
Bill Dennis
Active Member
Joined: 17 Aug 2007 Posts: 562 Location: Iowa, USA
|
|
|
|
You are probably in an LPAR with limited memory (512M is small!) and also probably limited on amount of CPU available. Someone designed the test environment to be limited capacity. If it restricts productivity, speak to your managers. |
|
Back to top |
|
|
|