How can I improve refresh time for Events in the AppManager Control Center Console? (NETIQKB71110)

  • 7771110
  • 20-Jun-2008
  • 06-Jul-2011

Environment

NetIQ AppManager 7.0.x
NetIQ Control Center 7.0.x

Situation

How can I improve refresh time for Events in the AppManager Control Center Console?

Resolution

Important SQL Configuration Information:

AppManager 7.0 and above has a greater reliance on performance tuning of the tempdb. The following Microsoft article contains more information on the internal working of SQL Server that can lead to the issue:

http://support.microsoft.com/kb/328551/

NetIQ HIGHLY recommends implementing this method in order to further increase the speed and efficiency of Microsoft SQL on servers hosting high-use databases, such as the AppManager Repository (QDB), Control Center Database (NQCCDB) and the AppManager Performance Profiler Database (AMPPDB).

 

Additional Information

There are changes internally within some of the routines in the QDB that make more use of the tempdb than previously. As a result, it is more important now in AM7 to tune SQL properly. The Microsoft Article above is vital for multi-proc servers.

 

Ideally, each file should be on a separate spindle. But often this is not possible. However, the QDB data and log files should be on separate spindles with the log file roughly half the size of the data file.

 

The tempdb should at least be moved from the default location (which will be in the Program files\Microsoft SQL Server\MSSQL.<x>\MSSQL\Data directory). The tempdb file(s) should be on separate spindles to the tempdb log file.

 

In general, data files should be located on a RAID 1 (mirrored) drive, and transaction logs on a RAID 0 (striped) drive. RAID 5 is not a very good substitute.


It would be considered generally good practice to size the tempdb as follows:-

      • Size the data file to be at least one third the size of the largest database (ie if the QDB is 30 Gb, make the tempdb data file 10 Gb)
      • Split the tempdb data file according to the number of processors (ie if there are 8 processors and the tempdb is 10 Gb, then split that into 8 2 Gb files)
      • Make the transaction log for the tempdb on half the size of the data file(s)

It should also be ensured that the transaction log for the QDB is about half the size of the data file for the QDB.

 

Where possible, each file (data/log for the QDB and data/log for the tempdb) should be placed on separate drives.

 

Please contact Microsoft Technical Support if you need assistance with implementing this recommended configuration.


Event Synchronisation Performance

There are three primary recommendations for improving Event Synchronization Performance in Control Center:

Recommendation 1 ? Increase Synchronisation Threads

The first recommendation is to increase the number of threads used to perform synchronisation from 10 to 20.

 

Having more threads running the synchronisation process quite simply means more ?entities? will be processed at any one time. However, testing has revealed that increasing the threads above 20 does not provide a linear change to synchronisation times, and begins to introduce an overhead in SQL Server.  Therefore, it is recommended that customers not increase the number of Synchronization Threads above 20.


To increase the threads used requires the following T-SQL statement to be executed against the Control Center Database:

update dbo.Preference set Value = 20 where PreferenceName = 'SyncBufferTaskCount'

After running this T-SQL statement, the NetIQ Command Queue Service should be restarted in order for it to take effect.  This will only have a limited effect on performance due to the number of entities that must be synchronised.  A larger effect will be seen by reducing the number of entities.

 


Recommendation 2 ? Remove Unnecessary Event Views


The second recommendation is to consider if those Management Groups that are in place to deploy Policies actually need an Event View at all.

 

If it is the case that a Management Group is used to determine where to deploy Policy Jobs, and another Management Group is used to define which events are viewed, it might not be a requirement for the Management Group used for Policy definition to have an Event View at all.

 

In this case, removing the cache for the Event View from those Management Groups that do not require it would potentially significantly reduce the number of entities processed and thus the overall synchronisation time. To do this, change the Cache properties of each Management Group that does not require events to match the settings shown in the attached file (see EventCacheSettings.jpg)

 

The setting for ?No jobs? in the above attached file is optional.

 


Recommendation 3 ? Remove ?Empty? Repositories

 

The third recommendation is to remove all ?unused? Repositories from Control Center.  Any Repository that is added to Control Center will be included in the Event Refresh process, whether that Repository is actually in use in Control Center or not.  Repositories should only be introduced to Control Center as they are needed.  If all Repositories that are connected to Control Center are in use in Control Center, this recommendation should not be used.

Additional Information

Formerly known as NETIQKB71110

NETIQKB71110.jpg