on 09-06-2012 2:24 PM
Hi experts,
Recently we began noticing performance drops in our MII system. I have trouble locating the bottleneck and would like your input.I have read lots of articles and discussions about best practices etc.
The webpages periodically call transactions which all together seem to take a lot of time to load. When calling a transaction on button click it often takes 5-15 seconds before the transaction executes. This is no Issue in QAS.
Our Production server has 24 logical cores and 28 MB RAM. The memory reaches amounts of up to 80% but the CPU's don't go much higher than 30%.
It does not seem to be a network related issue either.
When i check the schedular most transactions run under 1 second. These only include the sceduled transactions, not the ones called from applets.
Is there any way to get a list of transactions and how much time they / how often they run? Is there anyway to monitor the database (SQLSERVER 2008) to check if this might be the bottleneck? Any other paths I should look into?
Any help is greatly appreciated as this is becomming a major issue.
Thank you,
Philippe
These are the steps I would follow. Its off the top of my head, Its been a while since I had to do it, and I don't have all the docs handy and things change across Netweaver/MII versions anyway but maybe it will point you in the right direction. I think there is a document on the SDN concerning MII performance. Maybe other can chime in with good links to some details. I do still strongly recommend bringing someone in that has some exp though. In a day or two a good resource (even remotely) should be able to solve the issue assuming it is a Netweaver tuning problem.
Regards,
Christian
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello,
I'm joining this discussion, as we also encounter MII performance issues on our production system...
We have a lot (really a lot!) of transactions exposed as web services which are called by our shop floor system (every production machines have a POD to report production, downtimes, etc in SAP ME via MII transaction)
But for some reasons, the performance is degrading continously... What we noticed is that when we restart the server (for example after a new installation), the performance are really good, but starts quickly to degrade. We suspect a kind of memory problem (cache ? Wrong garbage policy ?), but we are a little bit blind and don't really know where to check. I'm not a MII or NW expert, only the developper.
Any suggestions others than what's above ? Is there a way to clean all caches without restarting ?
We are using MII 12.1.9.2 and NW 7.11.6.0
Thanks
Olivier
Hello,
Have you considered implementing Wily Introscope for your MII system. The good point is you can monitor and check historical values of different KPIs in a nice dashboard e.g. Garbage Collection, memory utlization, CPU, thread utilization etc. Based on your observation you can identify the root cause and tune.
Thanks,
Samik
Hello,
We had Wily in place for a while, but doesn't really help us, and we decided to stop this service as it was also consumming a lot of resources...
I was able to optimize some top-consumming BLS applying this process :
This helped a little bit, but it seems anyway we reached the limit of the servers, and we are now investigating with our infrasctructure team to upgrade the servers...
Regards
Olivier
Hello Olivier, just out of the blue, I'd check 3-4 things which by quickly looking were not in the list of things:
- the size of your tables: both custom and nw tables, and ensuring that the XMII_TRANSACTION table is not in the millions of records, which could adversely affect the performances. Just setup a job deleting those record periodically (limit the numbers of record deleted though) or wait for a scheduled production downtime.
- if you call your transaction through url with the runner, then the logging should be defined in the url add &LogType=Error, this could cause the XMII_TRANSACTION Table to log data, and this parameter is not overridden by the default in MII configuration page.
- Check by doing a join between XMII_FILE and XMII_PATH that you don't have a folder filling up by some test files which are consistently created by the developer for debug
- Ensure that any file stored in the database are regularly cleaned, if they are required. A past behavior of MII was to load all files stored in the database into the memory in order to increase the performance. This has an adverse effect only 20 mins after a reboot of the server because it is not immediately triggered and would cause you memory to fill up quite quickly.
Hope it helps,
Cheers,
Arnaud
Hi,
These are two excelent docs to read which can guide you in a direction in order to trouble shoot :
Regards
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello,
Thanks for the docs.
I have question about "usage monitoring"... I guess this will gather statistics like if we were using the "&LogStatisticsToDb=true, right ? If yes, how can we then retrieve those data's and how to clean them when finished ? I want first to try on our consolidation system before trying on live server. I think this will help us to identify the bottleneck in our top-consummer BLS.
Regards
Olivier
Hello,
I still don't know how to gather the stats info from the table... I have access to all XMII_STATS* table, I see some entries in the FILE table, nothing in the TRX table, which looks to me the more interesting... Any idea why ?
I also used the BLSManager&Mode=Stats&ID=xxx url based on a transaction shown in the transaction manager in the period I did activate the usage monitoring, but nothing displayed too...
I'm a little bit lost with this feature and to understand how it works...
Thanks
Olivier
Hi Philippe,
What version of MII are you running? Please include the SP and build. Also what version and SP of NW are you running if you are 12.X?
Thanks, Mike
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Philippe,
The simplest way would be to ask the NW admin or BASIS person. I think you can also get it from the NWA screen, but you may not have access to it (security folks tend to be a bit restrictive).
Do you have logging turned on? You may want to change the settings as it can fill up the database eventually.
Look at your Transaction Manager settings which record execution records of each transaction. If you have lots of logging of large xml documents, that can also add to storage.
You may also want your DBA to take a look at the underlying NW database and in particular the MII related tables.
Regards, Mike
Hi Michael,
Unfortunatly there is no BASIS person currently on the project. We changed the logging policy last week from 'ALWAYS' to 'ONERROR'. This helped some bit, especially the RAM usage lowered.
I am the DBA for the MII related tables. I don't see a performance issue when running queries directly on the database.
I think the problem might be the amount of transactions that run at the same time.We have around 20 workcenters each with a webpage. Every 15 seconds these pages refresh multiple data in grids. This can easily result in some 100 transactions run every minute.
These are very small transactions though that take under 100ms to run normally.
Hi Philippe,
Have you reviewed these documents? http://wiki.sdn.sap.com/wiki/display/xMII/Guidelines
This should have the performance checks listed for you. Look for the most recent change by Mike Solwecki, then the best practices.
Regards, Mike
Message was edited by: Michael Appleby
User | Count |
---|---|
9 | |
7 | |
3 | |
2 | |
2 | |
2 | |
1 | |
1 | |
1 | |
1 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.