Notifications

1847 views

MID Server Performance Reports | How to gauge the performance of deployed MID Servers


Introduction


The following reports are included by default starting with the Eureka release. Users on earlier releases can create these reports from the Reports module. For information on creating reports, see Reporting.

 

MID Server JVM memory


By default, the maximum memory allocated for each MID Server Java application is 512MB.  This value can be configured within the wrapper configuration by the property wrapper.java.memory.  For details on how to modify this value for a specific build version, see MID Server Configuration.

The amount of memory available to the MID Server is a function of the size of data to process and the number of concurrent jobs to process. Memory metrics are reported back from each MID Server and stored within the ECC Agent Memory Metric [ecc_agent_memory_metric] table on the instance. To gauge if the MID Server has been resourced with enough memory, one useful metric is the max used percentage (max_used_pct).  This metric is a percentage of the max used memory over the total available memory.  Trending this over time provides a visualization of how much memory is needed by the MID Server.

The example shows what the report may look like:

 

Each line represents a MID Server.  In this example, each MID Server uses, on average, 10% of its max memory.  We recommend that the max used percentage stays below 80%. This allows the MID Server to handle a majority of its work load as well as the occasional unexpected spikes. Here are recommendations to reduce the memory percentage:

  • Allocate more max memory to the MID Server. For more information, see MID Server Configuration.
  • Add additional MID Servers to share the workload. For more information, see the MID Server Clustering.
  • Reduce the amount of concurrent processing for the MID Server. This includes segmenting IP Address ranges into smaller segments for a Discovery Schedule or loading smaller segments of data within an import job.

 

MID Server host CPU usage


The MID Server also reports the CPU usage of its host. As the workload of a MID Server increases, so does the requirements for host resources. In addition to memory consumption, CPU consumption should also be monitored. CPU metrics are reported back from each MID Server about every ten minutes and stored within the ECC Agent Scalar Metric [ecc_agent_scalar_metric] table on the instance. Trending the daily avarage on CPU usage helps illustrate the amount of CPU processing that the MID Server host consumes.

The example shows what the report may look like:

Each line represents a MID Server Host. In this example, spikes in CPU usage can be seen that matche scheduled Discovery runs.  We recommend that the percentage stays below 80%.  If the percentage exceeds this amount, the recommendations are to reduce the activity on the host or migrate the MID Server to a new host with more available resources.

Note: MID Servers deployed on the same host, will report the same CPU usage.

 

Article Information

Last Updated:2017-08-07 15:14:50
Published:2014-08-08