|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
 |
|
Have you encountered lesser throughput with IIB v 10.0.0.9 w |
« View previous topic :: View next topic » |
Author |
Message
|
iibanalyst |
Posted: Thu May 31, 2018 2:27 am Post subject: Have you encountered lesser throughput with IIB v 10.0.0.9 w |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
Hi,
While migrating an Application from IIB v9.0.0.5 to IIB v.10.0.0.9, we have encountered a performance problem. We have raised a PMR with IBM. We are just distracted and delayed by IBM consultant instead of looking at the performance problem reported on IIB v10.0.0.9.
We have a simple application with a single message flow with the following nodes: MQ Input --> ESQL Compute (ODBC INSERT INTO DB2 Table)
No special logic used in the ESQL Compute node other than INSERT INTO STATEMENT to DB2 table.
On IIB v9.0.0.5, we have got a throughput of 3327 message per second with 19 additional instances. On IIB v10.0.09, we have got a throughput of 1253 message per second with 19 additional instances.
Same ODBC data source and DB2 table is used for testing the solution on IIB v10.0.0.9 and 9.0.0.5.
Have any of you encountered such huge performance problem with IIB 10.0.0.9 either after migrating from earlier version or on an application newly created on IIB 10.x ?
The difference in performance is huge we can't afford to have this drop in performance in an environment where we process few million MQ messages a day.
Please help !!! Advance thanks !!
Please message me for project interchange files, ddls and few other traces I have gathered for an IBM PMR. |
|
Back to top |
|
 |
abhi_thri |
Posted: Thu May 31, 2018 3:03 am Post subject: |
|
|
 Knight
Joined: 17 Jul 2017 Posts: 516 Location: UK
|
hi...usually for performance issues there could be multiple factors coming into play.
You haven't mentioned the server memory/kernel params (assuming it is a unix platform). Try comparing the hardware, memory, kernel, user file/process limits etc between the servers. We did face a performance degradation issue while migrating from Win 6.x Broker to Linux IIB 9.0.0.6 but at the end it was due to the Redhat parms used to build the Linux server (the kernel autotune param was turned off by the build team, i think it is on by default on Redhat vanilla build.)
If the backend DB2 database hasn't changed you can rule out any database issues, having said that I guess the v10 will be using a different DB2 driver version? Also is it worth verifying that there are no other (or atleast similar) database load in regards to other parallel apps using the same database while running the performance tests.
Also, in v10 have you tried increasing the no. of additional instances gradually to see whether the throughput is peaking at a lower no. and then falling off. |
|
Back to top |
|
 |
fjb_saper |
Posted: Thu May 31, 2018 2:18 pm Post subject: |
|
|
 Grand High Poobah
Joined: 18 Nov 2003 Posts: 20756 Location: LI,NY
|
As this is V10 vs v9 of IIB you did not specify if your IIB10 was still using a local queue manager. The move from a local queue manager (server bindings) to a client connection could account for your drop in performance.  _________________ MQ & Broker admin |
|
Back to top |
|
 |
iibanalyst |
Posted: Mon Jun 04, 2018 4:37 pm Post subject: The source code is here in DWforum as attachments |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
|
Back to top |
|
 |
iibanalyst |
Posted: Mon Jun 04, 2018 4:47 pm Post subject: Testing process and Environment used |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
The entire load test was executed on the same machine with exact same configuration (CPU, RAM, DB2 Version, Table, disks pace). I have listed the specification of the machine and the software version below.
The Application/Message flow was created in IIB v9.0 to just insert into DB2 table using ESQL Compute node (via ODBC). There is no special logic used in the ESQL Compute Node other than INSERT INTO statements. And the DB2 table do not have any unique key constraints on the INPUT VALUES.
Load test was executed by following the below steps for both IIB v9 and IIB v10:
1. Load 200,000 messages onto the INPUT QUEUE
2. Start message flow statistics
3. Start the message flow
4. View the message flow statistics
Those message throughputs were extracted from Message flow statistics.
Specification of the machine and software versions
• DB2 Server was running locally and its version is DB2/NT64 10.5.8 • IIB 9.0.0.5 • IIB 10.0.0.9 • MQ 7.5.0 • 4 Core CPU (Intel Core 2 Duo) • Windows 7 Professional 64 bit • RAM: 8GB |
|
Back to top |
|
 |
abhi_thri |
Posted: Tue Jun 05, 2018 12:22 am Post subject: |
|
|
 Knight
Joined: 17 Jul 2017 Posts: 516 Location: UK
|
As you've rule out most of the external factors (as the server, database etc are the same) I suggest to raise this issue with IBM via a PMR and see how it goes (and please do update us as well). |
|
Back to top |
|
 |
timber |
Posted: Tue Jun 05, 2018 2:02 am Post subject: |
|
|
 Grand Master
Joined: 25 Aug 2015 Posts: 1292
|
I would try varying the test.
- Replace the INSERT with a SELECT and re-test (in case it's something to do with transaction management)
- Replace the INSERT with some trivial piece of ESQL computation ( to eliminate the database entirely)
Obviously, in each case you need to test on both versions of IIB. |
|
Back to top |
|
 |
iibanalyst |
Posted: Tue Jun 05, 2018 4:07 pm Post subject: @abhi_thri - a PMR is raised with IBM |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
Hi @abhi_thri,
Thanks for your message, I have raised a PMR with IBM. IBM is working on it. It is a severity1 but it has taken more than 35 days so far. Not a positive response from IBM received. I was posting here to see if anyone has faced a similar problem. Direct users of IIB 10 may not find the differences because they never know what is the potential behind IIB v9/WMB 8/WMB 6/WMB 7.
Thanks
iibanalyst |
|
Back to top |
|
 |
iibanalyst |
Posted: Tue Jun 05, 2018 4:11 pm Post subject: @timber I will give it a go |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
Hi @timber,
I will give a go with your suggestion.
Thanks,
iibanalyst |
|
Back to top |
|
 |
abhi_thri |
Posted: Tue Jun 05, 2018 10:56 pm Post subject: Re: @abhi_thri - a PMR is raised with IBM |
|
|
 Knight
Joined: 17 Jul 2017 Posts: 516 Location: UK
|
iibanalyst wrote: |
Hi @abhi_thri,
Thanks for your message, I have raised a PMR with IBM. IBM is working on it. It is a severity1 but it has taken more than 35 days so far. Not a positive response from IBM received. I was posting here to see if anyone has faced a similar problem. Direct users of IIB 10 may not find the differences because they never know what is the potential behind IIB v9/WMB 8/WMB 6/WMB 7.
Thanks
iibanalyst |
Oh, you already mentioned that a PMR is in place in your first post, didn't notice that. Must admit, from what I've observed IBM response times for PMRs seems to be getting worse over the past few years.
Quote: |
• DB2 Server was running locally and its version is DB2/NT64 10.5.8 • IIB 9.0.0.5 • IIB 10.0.0.9 • MQ 7.5.0 • 4 Core CPU (Intel Core 2 Duo) • Windows 7 Professional 64 bit • RAM: 8GB |
Is this your local desktop machine Or do you use Windows in all environments including Prod? If you have a Linux estate it may be worth installing v10 on Linux (with a relatively high spec hardware) and see that makes any difference.
Also, are you noticing any significance difference in CPU/memory usage during the performance runs b/w v9 and v10.
Also, have you tried running the tests with the flow statistics turned off (guess you should still be able to measure the throughput by looking at the MQ put/get queue stats and database insert/update timestamps) |
|
Back to top |
|
 |
iibanalyst |
Posted: Wed Jun 06, 2018 5:04 pm Post subject: @Performance values on Window and AIX |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
I have done further testing on (IIB) AIX machine with DB2 on AIX. Below are the latest reports which include the Windows and AIX numbers.
IIB Version and Context, "IIB OS AIX or Windows 7","DB2 Local or AIX ? ","Number of messages","Additional Instances",Start time,End time,"Processing time","Processing time in seconds", Message Rate (messages/second)
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, AIX, 199990, 19, 12:14:25, 12:16:23, 0:01:58, 118, 1695
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, LOCAL, 50000, 19, 15:16:30, 15:16:57, 0:00:27, 27, 1852
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, LOCAL, 50000, 9, 15:20:50, 15:21:24, 0:00:34, 34, 1471
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, LOCAL, 50000, 11, 15:24:55, 15:25:20, 0:00:25, 25, 2000
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, LOCAL, 50000, 13, 15:34:42, 15:35:07, 0:00:25, 25, 2000
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, LOCAL, 50000, 10, 15:40:39, 15:41:06, 0:00:27, 27, 1852
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, LOCAL, 50000, 13, 15:45:59, 15:46:26, 0:00:27, 27, 1852
"IIB 10.0.0.9 CLIENT CONNECTION", Windows 7, LOCAL, 99990, 11, 10:50:59, 10:51:55, 0:00:56, 56, 1786
"IIB 9.0.0.5 LOCAL CONNECTION", Windows 7, AIX, 100000, 11, 14:15:47, 14:16:32, 0:00:45, 45, 2223
"IIB 9.0.0.5 LOCAL CONNECTION", Windows 7, AIX, 100000, 19, 14:22:17, 14:22:48, 0:00:31, 31, 3226
"IIB 9.0.0.5 LOCAL CONNECTION", Windows 7, LOCAL, 199990, 19, 14:50:54, 14:52:24, 0:01:30, 90, 2223
"IIB 9.0.0.5 LOCAL CONNECTION", Windows 7, LOCAL, 50000, 19, 15:07:29, 15:07:46, 0:00:17, 17, 2942
"IIB 9.0.0.5 LOCAL CONNECTION", Windows 7, LOCAL, 50000, 11, 15:54:46, 15:55:03, 0:00:17, 17, 2942
"IIB 9.0.0.5 LOCAL CONNECTION", Windows 7, LOCAL, 199990, 11, 16:33:15, 16:34:13, 0:00:58, 58, 3449
"IIB 10.0.0.9 CLIENT CONNECTION", AIX, AIX, 100000, 19, 17:59:04, 18:01:26, 0:02:22, 142, 704
"IIB 10.0.0.9 CLIENT CONNECTION", AIX, AIX, 100000, 11, 18:08:10, 18:09:58, 0:01:48, 108, 926
"IIB 10.0.0.9 CLIENT CONNECTION", AIX, AIX, 100000, 13, 18:26:45, 18:28:21, 0:01:36, 96, 1042
"IIB 9.0.0.5 LOCAL CONNECTION", AIX, AIX, 100000, 19, 18:44:03, 18:44:34, 0:00:31, 31, 3226
"IIB 9.0.0.5 LOCAL CONNECTION", AIX, AIX, 100000, 11, 18:51:31, 18:52:13, 0:00:42, 42, 2381
Last edited by iibanalyst on Wed Jun 06, 2018 6:03 pm; edited 1 time in total |
|
Back to top |
|
 |
iibanalyst |
Posted: Wed Jun 06, 2018 5:11 pm Post subject: Please try the solutions if you like to see yourself |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
|
Back to top |
|
 |
fjb_saper |
Posted: Wed Jun 06, 2018 5:58 pm Post subject: |
|
|
 Grand High Poobah
Joined: 18 Nov 2003 Posts: 20756 Location: LI,NY
|
Why am I not surprised that the local connection is way faster than the client connection???  _________________ MQ & Broker admin |
|
Back to top |
|
 |
iibanalyst |
Posted: Wed Jun 06, 2018 6:04 pm Post subject: |
|
|
Newbie
Joined: 31 May 2018 Posts: 9
|
fjb_saper wrote: |
Why am I not surprised that the local connection is way faster than the client connection???  |
** Corrected few rows to of type LOCAL because all the IIB 9.0.0.5 tests were done with LOCAL Queue Manager. |
|
Back to top |
|
 |
|
|
 |
|
Page 1 of 1 |
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|
|
|