| Author |
Message
|
| kenward |
 Posted: Mon May 09, 2011 7:01 am Post subject: Transactions per second or minute on Broker v7? |
 |
|
Apprentice
Joined: 01 Jun 2010
Posts: 41
Location: Detroit, MI
|
I've turned on both Resource and Message Flow stats but I can't seem to find any filters that include a transactions or messages per minute or second.
Anyone know an easy way to get flow invocations per time interval stats?
-Jeff |
|
| Back to top |
|
 |
| lancelotlinc |
| Posted: Mon May 09, 2011 9:00 am Post subject: |
|
|
Jedi Knight
Joined: 22 Mar 2010
Posts: 4941
Location: Bloomington, IL USA
|
To measure transactions per time interval, take a large group of messages, such as 1 million messages, and load them into a queue with the "get disabled" flag on. Note the time you turn the "get disable" flag off. Watch the queue depth. When the queue depth reaches zero, mark that time also. 1 million divided by the number of seconds elasped is the transactions per second.
For basic RHEL, you should be able to see around 10,000 TPS for a simple message flow. For a muscle AIX Power7, you should be able to see 500,000 TPS. For z/OS, roughly one-third of AIX or RHEL TPS because it takes 3 times more CPU time to process the same message on z/OS as it does on other platforms.
Here is an interesting paper to read:
http://www-01.ibm.com/support/docview.wss?uid=swg24025868
_________________
http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon May 09, 2011 9:05 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| lancelotlinc wrote: |
| To measure transactions per time interval, take a large group of messages, such as 1 million messages, and load them into a queue with the "get disabled" flag on. Note the time you turn the "get disable" flag off. Watch the queue depth. When the queue depth reaches zero, mark that time also. 1 million divided by the number of seconds elasped is the transactions per second. |
This assumes the OP is trying to measure pure performance, not resources used and/or messages processed during a given business period. For instance, the business could be asking how many customer detail enquiries were made by users & processed by broker per hour.
| lancelotlinc wrote: |
| For basic RHEL, you should be able to see around 10,000 TPS for a simple message flow. For a muscle AIX Power7, you should be able to see 500,000 TPS. For z/OS, roughly one-third of AIX or RHEL TPS because it takes 3 times more CPU time to process the same message on z/OS as it does on other platforms. |
I suspect that if the OP is trying to enable stats on a running flow, the purchasing decision regarding platform was made some time ago...
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| mqjeff |
| Posted: Mon May 09, 2011 9:10 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
What do you consider to be a "transaction"?
A single execution of a message flow?
The statistics data provides the start and stop times of the statistics interval, which you can use the calculate the number of seconds covered. You can then use the number of messages processed at the message flow level to calculate the "transactions" per second. |
|
| Back to top |
|
|
| smdavies99 |
| Posted: Mon May 09, 2011 9:22 am Post subject: |
|
|
Jedi Council
Joined: 10 Feb 2003
Posts: 6076
Location: Somewhere over the Rainbow this side of Never-never land.
|
IT hasn't been specifically mentioned here but the Throughput does depend upon the overall system performance.
This is not only the CPU but the memory subsystem, the Storage subsystem and possibly the Networking subsystem.
Naturally, you won't see the same TPS figure on a single core Intel Atom as you would on an 6 core AMD Magny cours CPU. That is why the IBM Performance reports (available as support packs from the link at the top of the page) do mention the H/W being used.
One system I benchmarked went from 41 to 58 TPS when we changed the way the network connection was established. This was on a Windows Server 2003 system with a remote SQL-Server DB. This improvement was done by changing disabling the Auto-Negociate in the network interface.
_________________
WMQ User since 1999
MQSI/WBI/WMB/'Thingy' User since 2002
Linux user since 1995
Every time you reinvent the wheel the more square it gets (anon). If in doubt think and investigate before you ask silly questions. |
|
| Back to top |
|
|
| kenward |
| Posted: Mon May 09, 2011 9:22 am Post subject: |
|
|
Apprentice
Joined: 01 Jun 2010
Posts: 41
Location: Detroit, MI
|
Thanks Jeff, I'll try that.
-Jeff |
|
| Back to top |
|
|
| lancelotlinc |
| Posted: Mon May 09, 2011 9:33 am Post subject: |
|
|
Jedi Knight
Joined: 22 Mar 2010
Posts: 4941
Location: Bloomington, IL USA
|
| mqjeff wrote: |
| What do you consider to be a "transaction"? |
My test cases were similar to the ones identified here:
http://www-01.ibm.com/support/docview.wss?uid=swg24025868
On page 59, CPU time per message is about 1.5 ms for 2,000 byte message. On z/OS, performance for similar test case was twice that per message for 1,000 byte messages.
Admittedly, its an apple to orange comparison, but anyone who has worked on WMB for any length of time would tell you their preference for a production deployment environment and very few would prefer z/OS as their first choice.
WMB 7 on Power7 AIX by far (multiples) has the best per message performance, in my opinion.
_________________
http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
| Back to top |
|
|
| lancelotlinc |
| Posted: Mon May 09, 2011 10:34 am Post subject: |
|
|
Jedi Knight
Joined: 22 Mar 2010
Posts: 4941
Location: Bloomington, IL USA
|
| Vitor wrote: |
| I suspect that if the OP is trying to enable stats on a running flow, the purchasing decision regarding platform was made some time ago... |
Did you mean in the 1960s?
_________________
http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon May 09, 2011 10:38 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| lancelotlinc wrote: |
| Vitor wrote: |
| I suspect that if the OP is trying to enable stats on a running flow, the purchasing decision regarding platform was made some time ago... |
Did you mean in the 1960s? |
No, I meant long enough ago that the hardware has been installed, the software stack set up, broker installed and a flow developed. It's probably a bit late for them to slap their foreheads and say, "hang on, this is all wrong, let's go buy AIX on Power 7".
The OP has given no indication of the OS in use and could be using a 21st Century version of Windows for all you know. Nowhere in this thread has anyone except you mentioned z/OS (which I'm guessing is where the 1960s jibe is coming from).
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| lancelotlinc |
| Posted: Mon May 09, 2011 10:55 am Post subject: |
|
|
Jedi Knight
Joined: 22 Mar 2010
Posts: 4941
Location: Bloomington, IL USA
|
@Vitor
I was merely providing target TPS rates that people could use as ballpark for determining sanity-check on the values they see. I did not indicate that anyone should replatform their installation.
A little warm today, Vitor, eh?
_________________
http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon May 09, 2011 11:10 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| lancelotlinc wrote: |
| A little warm today, Vitor, eh? |
Too warm in here for me these days. No matter who asks.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| joebuckeye |
| Posted: Mon May 09, 2011 11:56 am Post subject: |
|
|
Partisan
Joined: 24 Aug 2007
Posts: 365
Location: Columbus, OH
|
| Quote: |
| I was merely providing target TPS rates that people could use as ballpark for determining sanity-check on the values they see. |
But the numbers you give are specific to your flow. How can they be of use to other people?
I've seen you throw these numbers out before and wondered why you think they mean anything to anyone else.
Everybody's transformation needs are different, same with auditing or logging, or anything else that companies want their message flows to do.
Unless you give out a lot of details about what your flow is doing and how it is doing it the numbers you give out are useless to other people. |
|
| Back to top |
|
|
| lancelotlinc |
| Posted: Mon May 09, 2011 12:14 pm Post subject: |
|
|
Jedi Knight
Joined: 22 Mar 2010
Posts: 4941
Location: Bloomington, IL USA
|
| joebuckeye wrote: |
| Quote: |
| I was merely providing target TPS rates that people could use as ballpark for determining sanity-check on the values they see. |
But the numbers you give are specific to your flow. How can they be of use to other people?
I've seen you throw these numbers out before and wondered why you think they mean anything to anyone else.
Everybody's transformation needs are different, same with auditing or logging, or anything else that companies want their message flows to do.
Unless you give out a lot of details about what your flow is doing and how it is doing it the numbers you give out are useless to other people. |
My numbers are based on the test cases here:
http://www-01.ibm.com/support/docview.wss?uid=swg24025868
They mean something when comparing a unit of work in different environments.
You are correct, Joe, to get a true comparison, you need the same message flow in each environment you want to compare.
I was not publicizing a concrete benchmark, only a general apple to orange comparison.
If a simple message flow on RHEL can only do 10 TPS, something is wrong. MQInput -> Compute -> MQOutput on RHEL should be relatively in the 10,000 TPS range give or take a few thousand.
Performance is always relative and your mileage will be different than mine. If we both bought the same model and make of car, even if all things are identical on the car, there will be some variance. But if I getting 34 MPG and you getting 3 MPG, you need to take your car in for service.
_________________
http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
| Back to top |
|
|
|
|
