+

Search Tips   |   Advanced Search

Performance tuning

Learn about performance tuning for the MPF Operational Analytics.


MobileFirst throughput-tuning

We can tune the amount of data that is held on the MobileFirst Server to balance the risk of data loss against the risk of overloading the MPF Operational Analytics. When Analytics data is sent from the client, it is bundled together and placed in a queue on the MobileFirst Server. When a queue is filled up on the MobileFirst Server, it posts that data to the Analytics Platform. Then, the queue is emptied.

Use the following two parameters in the JNDI configuration for throughput-tuning:

The wl.analytics.queues parameter determines the maximum number of queues that the MobileFirst Server holds in memory. If all of the queues fill up before they post to the Analytics Platform, the MobileFirst Server drops the most recently received data.

The wl.analytics.queue.size parameter is the number of individual elements that each queue can hold. Adjustment of these parameters affects:

The following image shows the data accumulation on the MobileFirst Server:

The number of individual analytics events that the server holds at one time is wl.analytics.queues * wl.analytics.queue.size. Take this fact into consideration when you define these two parameters. If we set them too low, large amounts of analytics data can be dropped if the server is unusually busy. If we set them too high, too much memory can be used on the MobileFirst Server.


Java virtual machine (JVM) swapping

The underlying technology used by the MPF Operational Analytics is called Elasticsearch. Elasticsearch performs poorly when the JVM starts swapping. To ensure that the JVM never swaps, the following JNDI property can be set to true:


Parent topic: Production deployment and clustering