| Author |
Message
|
| SOAIntegration |
 Posted: Mon Jul 09, 2012 12:03 pm Post subject: Increase of MIPS with Broker on Mainframes |
 |
|
Newbie
Joined: 07 Dec 2010
Posts: 7
|
Hello,
we have message Broker hosted on mainframes and we have different kinds of flows which use simple ESQL and few of them using Java compute node with complex java transformations. in the past few months we are seeing in lot of increase of MIPS which is causing too much of cost for maintenance.
Can someone tell me what are different factors that cause increase in MIPS. I mean by using any specific nodes or code constructs which increase MIPS or DB etc....
any pointers to optimize the code and decrease the MIPS count is required. thanks in advance.
regards
UD |
|
| Back to top |
|
 |
| lancelotlinc |
| Posted: Mon Jul 09, 2012 12:30 pm Post subject: |
|
|
Jedi Knight
Joined: 22 Mar 2010
Posts: 4941
Location: Bloomington, IL USA
|
The mainframe is the least efficient of any WMB platform. Of the Broker nodes, ESQL is slightly more efficient than JCN.
_________________
http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
| Back to top |
|
|
| mqjeff |
| Posted: Mon Jul 09, 2012 12:37 pm Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
| Why haven't you already read the performance report? |
|
| Back to top |
|
|
| Mr Butcher |
| Posted: Mon Jul 09, 2012 11:08 pm Post subject: |
|
|
Padawan
Joined: 23 May 2005
Posts: 1716
|
check for updates you done close before you see the increase in MIPS. did you change something in your db2 environment? java? did you check your storage assignments for execution groups, especially when running java stuff? maybe switch on storage reports...
do you have broker accouting switched on? maybe this gives a hint at wich area the increase happens.
another option is to involve ibm....
good luck
stefan
_________________
Regards, Butcher |
|
| Back to top |
|
|
| zpat |
| Posted: Mon Jul 09, 2012 11:43 pm Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
| I'm a bit out of touch with z/OS (unfortunately) but I suspect using one of those z-Series Java engines would make a huge difference and reduce the use of standard MIPS. |
|
| Back to top |
|
|
| Maximreality |
| Posted: Fri Aug 31, 2012 4:45 am Post subject: Re: Increase of MIPS with Broker on Mainframes |
|
|
Acolyte
Joined: 04 Jun 2004
Posts: 65
Location: Copenhagen
|
...as zpat is mentioning, tak a look at zAAP as this might be relevant for you, as this offloads java processing and reduces MIPS
http://en.wikipedia.org/wiki/Z_Application_Assist_Processor
You might have a mainframe without zAAP og it might be disabled ?
If you have heavy db2 usage zIIP might be relevant as well
http://en.wikipedia.org/wiki/ZIIP
Using the MRM parser is heavy in CPU usage - with tds/cwf messages using substring to build your own "parser" will in most cases use less CPU than the MRM parser.
Always use the XMLNSC parser for XML as this uses less CPU than XML or XMLNS
In general you can optimize our messageflows with having fewer compute nodes - as everytime you enter a computenode the broker starts to parse
Also look at the setting of the compute node - "Compute mode"
/Rasmus |
|
| Back to top |
|
|
| kimbert |
| Posted: Sat Sep 01, 2012 10:23 am Post subject: |
|
|
Jedi Council
Joined: 29 Jul 2003
Posts: 5542
Location: Southampton
|
| Quote: |
| Using the MRM parser is heavy in CPU usage - with tds/cwf messages using substring to build your own "parser" will in most cases use less CPU than the MRM parser. |
At this point we do not even know whether the OP is using the MRM parser. Your point may be factually accurate, but it misses the point. If you write your own parsing logic you will incur maintenance costs - and they will not be small.
v8 users may want to explore the performance benefits of using DFDL instead of MRM.
| Quote: |
| In general you can optimize our messageflows with having fewer compute nodes - as everytime you enter a computenode the broker starts to parse |
Careful there! Reducing the number of Compute nodes does reduce CPU usage, but that fact is unrelated to parsing. Message broker parses each message at most once - unless you have explicitly requested a complete parse using CREATE...PARSE in your ESQL. |
|
| Back to top |
|
|
| Maximreality |
| Posted: Sat Sep 01, 2012 12:24 pm Post subject: |
|
|
Acolyte
Joined: 04 Jun 2004
Posts: 65
Location: Copenhagen
|
| kimbert wrote: |
If you write your own parsing logic you will incur maintenance costs - and they will not be small.
|
...and that depends on several factors.
The guy was asking for ways to reduce CPU usage and what things in general uses a lot of CPU.
I know you guys from IBM are very happy with the MRM parser and doing things the right way...., but there is also a real world  (and sometimes a MIPS bill to be paid)
We can also discuss if the broker at all should run on a mainframe, because even when using zAAP etc. it will in most cases be more expensive than running it on a distributed platform from what i have experienced.
So it depends, do you really need the features of the mainframe ? |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Sat Sep 01, 2012 1:48 pm Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| lancelotlinc wrote: |
| The mainframe is the least efficient of any WMB platform. |
Care to explain this? Efficient is far too nebulous a term. What is inefficient? Dispatching? I/O? Concurrent processing?
Are you referring to WMQ on z/OS? zLinux?
Can you offer some facts or cite some references to back up your statement?
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Wed Sep 05, 2012 6:08 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| lancelotlinc wrote: |
| The mainframe is the least efficient of any WMB platform. |
This is one of those proclamations that needs some explanation.
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| Vitor |
| Posted: Wed Sep 05, 2012 6:17 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| bruce2359 wrote: |
| lancelotlinc wrote: |
| The mainframe is the least efficient of any WMB platform. |
This is one of those proclamations that needs some explanation. |
Or, as you indicated above, some context.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
|
|
