Ansicht
Dokumentation
06290 - Multiple instances on iSeries
Addresses (Business Address Services) ROGBILLS - Synchronize billing plansThis documentation is copyright by SAP AG.
Multiple instances on iSeries
Hello Franco,Let me start at the top.
You seem to be having a problem with database accesses. You are performing
allot of sequential reads compared with the direct reads. In addition, the
time for sequential reads is a little high which might indicate that you
have an opportunity to optimize database accesses with some indexes.
Change/commit time is very high which tells me that either ASP2 is part of a
raid-5 set or there are only 2 disk drives that are mirrored in the ASP.
There should be at least 4 disk drives (2 mirrored) in the ASP2.
I can't tell if the ST03 statistics are for dialog or for all work process
types. I'm making the assumption for now that this is for dialog.
The Av. DB req. time 1.310,9 ms should be 300 to 400 ms for sub second
response time. So, you are about 1000 ms too high there. This could partly
be caused by storage pool sizes, partly by bad SQL, and partly by index
opportunities. I see from the WRKSYSSTS that the machine pool is faulting.
This should not happen. I also see a high ration of pages to faults for
database which is good. You should probably add some memory to the machine
pool or clean up allot of the 2200 jobs in the system or both. Secondly,
make sure the base pool paging option is *calc. Third, set the activity
level in the *BASE pool to about 1.3 * the number of work processes in the
instance. Finally, make sure there aren't allot of FTP, telnet, HTTP, etc
servers running.
The Average wait time 100,8 ms is about 95ms too high. This was
probably caused by not enough work processes or many long running processes.
The Av. Roll i+w time 131,2 ms in conjunction with Roll wait time
26.569,7 s tells me that there is external RFC communications. There may be
an opportunity there to optimize some of that.
You need to add some disk space to the system ASP to account for the temp
storage that is being used. Probably 3-4 17GB drives would be about right.
This should keep disk utilization around 70% at peek. I can't tell from the
information here if you are having a problem with hot spots on the disks.
You would have to look at WRKDSKSTS and pieces of R/3 for that.
Best regards,
Alan
-----Original Message-----
From: Franco Tassone [mailto:franco.tassoneZe...]
Sent: Thursday, May 30, 2002 11:26 AM
To: DoNotReply@consolut.eu
Cc: Giampiero Bianchi; Riccardo De Luca
Subject: RE: Multiple instances on iSeries
Many thanks to all the folks who gave us their advices. We won't go on
installing multiple instances as our system is not a multi tier on and
never it will be.
We already did a tune up after our upgrade enlarging varios parameters and
buffers, but our performances are less than expected.
Our productive instance is running on an lpar of our iSeries 830, 6,5
processors (1,5 reserved for another lpar where dev and test instances
live) using about 8Gigs ram and about 350 Gigs database (on the IBM Ess).
As an example this is the workload registered yesterday, from ST03
SAP System F3C Date 29.05.2002 First record 00:00:01
Instance no. 01 Last record 23:59:59
Server S44G9031 Time elapsed 23:59:58
CPU time 71.359,5 s Database calls 22.716.903
Time elapsed 86.398,0 s Database requests 56.716.889
Direct reads 38.330.078
Dialog steps 175.700 Sequential reads 16.327.773
Changes 2.059.038
Average CPU time 406,1 ms
Av. RFC+CPIC time 15,9 ms Time per DB request 4,1 ms
Direct reads 0,7 ms
Av. response time 2.137,0 ms Sequential reads 10,1 ms
Average wait time 100,8 ms Changes and commits 18,6 ms
Average load time 51,8 ms
Av. Roll i+w time 131,2 ms Roll-in time 2.721,7 s
Av. DB req. time 1.310,9 ms Roll-out time 4.150,7 s
Av. enqueue time 0,7 ms Roll wait time 26.569,7 s
Roll-ins 253.802
Average bytes req. 152,7 kB Roll-outs 254.369
The following is the buffer and work processes configuration as show in
RZ10
ABAP programs 800000 KB Dialog 24
Nametab 45897 KB Update (V1) 5
Generic key tables 63477 KB Update (V2) 2
Single key tables 30000 KB Enqueue 1
CUA Information 10000 KB Background 6
Screens 976563 KB Spool 2
==> Initial / max. swap requirement = 11128 MB / 13036 MB
This is what's shown in the WRKSYSSTS screen:
Gestione dello stato del sistema S44G9031
30/05/02
17:20:55
% CPU in uso . . . . . . : 15,5 Memoria ausiliaria:
% capacità DB . . . . . : 9,8 ASP sistema . . . . . : 324,4
G
Tempo trascorso . . . . : 00:00:20 % ASP sistema utiliz. :
80,4780
Lavori nel sistema . . . : 2204 Totale . . . . . . . . : 359,5
G
% indirizzi permanenti . : 0,018 Att. non protetta util.: 32718
M
% indirizzi temporanei . : 0,068 Massima non protetta . : 53221
M
Immettere le modifiche (se consentite) e premere Invio.
Dim Dim
Lotto lotto riservata Max -----DB----- ---Non-DB---
sist. (M) (M) Attiv. Err. Pag. Err. Pag.
1 500,00 406,39 +++++ 0,1 0,1 12,9 54,5
2 8307,50 0,25 120 20,7 2176 116,7 180,4
3 10,00 0,00 5 0,0 0,0 0,0 0,0
4 20,00 0,00 10 0,0 0,0 2,4 4,7
Yesterday I enlarged the number of work processes because they often were
all busy. Unfortunatelly we also have a large number of "Z" programs using
too much the dialog processes I could see this in the transaction profile
of the workload analysis. With slight modifies many of them could rnu in
the background.
Unfortunatelly our PTF status is still at CUM1302510 and DB fix pack #5
(as described in APAR# II12833 02/06/02) and the kernel patch is 922. The
reason of the delay in mantaining the ptf & patches was due to system
availability reasons. A consultant of ours suggested to update the system
to the latest II12833 and to apply kernel patch 1124, because of our
performance issue
I know that non-optimized programs are an handicap for our system,
nevertheless do you think I'll see a prerformance improvements installing
the latest APAR + the latest kernel patch ?
Another question: our ESS shark uses raid-5 for the disk, so our ASP2
virtually is a raid-5, instead of a mirror: does this impact on the
performances too ?
Thanks to everybody.
----------------------------------------------------------------------------
Franco Tassone - (Emsar IS) tel.: +39-085-4442248
----------------------------------------------------------------------------
[Non-text portions of this message have been removed]
Have a look to our homepage at: http://www.consolut.net
DoNotReply@consolut.eu
Your use of consolut is subject to http://www.consolut.net
Durban Tours - Südafrika Safari
Fill RESBD Structure from EBP Component Structure CL_GUI_FRONTEND_SERVICES - Frontend Services
This documentation is copyright by SAP AG.
Length: 7727 Date: 20240601 Time: 224221 sap01-206 ( 3 ms )
