eG Monitoring
 

Measures reported by IfxMemGtMgrTest

The Memory Grant Manager (MGM) is a database server component that coordinates the use of memory, CPU virtual processors (VPs), disk I/O, and scan threads among decision-support queries. The MGM dynamically allocates the following resources for decision-support queries:

  • The number of scan threads started for each decision-support query

  • The number of threads that can be started for each query

  • The amount of memory in the virtual portion of database server shared memory that the query can reserve

When the database server has heavy OLTP use, and you find performance is degrading, you can use the MGM facilities to limit the memory resources granted to decision-support queries. During off-peak hours, you can designate a larger proportion of the memory resources to parallel processing, which achieves higher throughput for decision-support queries. When query execution is delayed due to the insufficient memory resources, the users may experience delayed response to their queries and also the performance of the database server degrades. To avoid such unpleasant situation, you should continuously monitor the MGM memory utilization to accurately find out whether the memory issues are for slow down in the query execution. This can be achieved using the IfxMemGtMgrTest test.

This test accurately reveals the details on MGM memory utilization of the database server.

Outputs of the test : One set of results for the Java application being monitored.

The measures made by this test are as follows:

Measurement Description Measurement Unit Interpretation
Total_memory Indicates the total size of the MGM memory. MB
Used_memory Indicates the amount of MGM memory that was utilized by the server. MB A high value of this measure indicates that the MGM memory is running out of space, you may need to allocate additional resources.
Free_memory Indicates the amount of MGM memory that is currently available for use. MB  
Used_percentage Indicates the percentage of MGM memory that was utilized by the server. Percent