And now back to everyone's (ir)regularly :-) scheduled programming...
Cheers, Martin
Martin Packer,
zChampion, Principal Systems Investigator,
Worldwide Cloud & Systems Performance, IBM
+44-7802-245-584
email: martin_pac...@uk.ibm.com
Twitter / Facebook IDs: MartinPacker
Blog:
https://www.ibm.com/developerworks/mydeveloperworks/blogs/MartinPacker
From: Kirk Wolf <k...@dovetail.com>
To: IBM-MAIN@LISTSERV.UA.EDU
Date: 01/04/2016 22:48
Subject: Re: Microprocessor Optimization Primer
Sent by: IBM Mainframe Discussion List <IBM-MAIN@LISTSERV.UA.EDU>
I'm curious, Andrew -
Have you looked at the com.ibm.jzos.fields package for field mapping of
records?
These can be made to use fancy support in the JVM for inline JIT
replacement of primitive access with machine instructions.
There is also some tools on the JZOS developerWorks site - the JZOS
RecordGenerator. This tool will allow you to read Assembler DSECTS or
Cobol copybooks and generate Java classes that use the runtime classes
from
com.ibm.jzos.fields.
See:
https://www.ibm.com/developerworks/community/groups/service/html/communityview?communityUuid=2acdb076-7582-45b5-93a5-781f90169bd3
Regards,
Kirk Wolf
Dovetailed Technologies
http://dovetail.com
PS> The JZOS RecordReader (for eligible data sets) uses a JNI library that
uses best-practices BSAM techniques for efficiently reading sequential
data
sets. The JNI interface is block-level, rather than record-level, which
helps a great deal.
The only part that you couldn't do yourself is to make the JNI library
zAPP/zIIP eligible. The JZOS JNI libraries *are* zAPP eligible, since
they are part of the IBM JRE.
On Thu, Mar 31, 2016 at 7:13 PM, Andrew Rowley
<and...@blackhillsoftware.com
wrote:
On 31/03/2016 23:34, David Crayford wrote:
Thanks for the info Andrew. We use the default GC options and our
workloads are both long running agents and batch jobs. I really do want
to
learn a lot more about how to tune the JVM for each specific workload.
The
SMF 30 records are useless. How do I find out how much time is spent in
JIT
overhead compared to actual processing? What tools do you use to
profile
your JVM apps?
You probably know more about it than I do. I am pretty new to Java on
z/OS, coming from C#. I did some basic GC testing by running the same
job
10 times and taking the average for different GC options. I tried
optthruput because it sounded like what I want for Java batch, but from
memory total CPU increased by 2 or 3 times. All the GC options I tried
changing either increased CPU or made no change, so I have just been
using
the defaults.
There is an option somewhere where you can see the time spent in GC, the
Java guys say that as a ROT if it is >5% try a different GC strategy.
There
are so many knobs you can tweak though you could spend an awfully long
time
measuring their effects and interactions.
I do know that the effects of allocation patterns are very different for
GC languages and languages like C/C++. In C++ allocating and freeing
short
lived objects is expensive because the memory needs to be tracked. In GC
languages short lived objects are cheap, it is objects that survive to
GC
that are expensive. This can make it a bit difficult to directly compare
performance because there might be different design decisions.
Most of my CPU time seems to be I/O related (processing SMF data). Even
removing all the reporting seems to make very little difference to the
CPU
consumed. I'm trying to work out if I can reduce the I/O overhead.
I don't know about JIT overhead - I think it is probably insignificant
because it should only compile the methods you actually invoke.
The FTS1 CPU is running CICS, IMS and that one Java batch job spikes
past
it.
Without other work for competition anything CPU bound will max out the
CPU, so that's not necessarily a negative - it just means that it isn't
waiting for other things. E.g. at one site the DBAs greatly increased
Adabas buffer space. CPU contention increased massively as virtually
nothing was waiting for I/O anymore. Caused a few problems, but it's a
good
problem to have. It just means that if it can only get 5% CPU instead of
50% it will run ~10 times as long. Total CPU time is a better measure
than
CPU%.
--
Andrew Rowley
Black Hill Software
+61 413 302 386
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number
741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
