Memory Usage Optimization

This chapter consists of the most important options, properties and settings for Iotellect Server memory usage optimization.

Models are the main instruments for data processing inside Iotellect Server. Often data calculation results are stored inside one or more model variables.

You can configure the Server In-memory Value Caching variable property to Keep While Enough Memory state. It will prevent memory overflow, but the overall performance of model variable read operations may decrease.

An Iotellect plugin as a JAR file is usually not very large, but can include third-party libraries with significant allocation activity. For example, the Unified Search plugin employs a third party library for context data indexing which in turn relies heavily on memory mapped files and can produce significant spikes of memory consumption.

To check which plugins are available on the server and to toggle them, open Server Configuration and navigate to the Active Plugins tab. Uncheck the Enabled flag to disable a given plugin. Restart the Iotellect server for the new configuration to take effect.

The Iotellect server uses both persistent and non-persistent caches for processing variable values during runtime. The non-persistent cache runs in-memory and uses some portion of the server memory allocated to the Java Virtual Machine. Almost all variable values are placed in cache after being read from the database to speed up subsequent readings. During normal server operation, Iotellect will typically start to utilize almost all allocated memory, as more and more variables are read from the database and placed in cache.

When the size of the cache reaches a certain threshold, a background process removes infrequently-used variables and reduces the cache size to an appropriate level. During normal server operation, this cleaning operation will cycle on and off as needed to ensure enough memory is available for normal system operation given the overall load.

Statistics regarding the operation of the Memory Clearing operation are displayed by the Server Information variable Memory Clearing, indicating how many times Memory Clearing has run, when it last started, and how long the most recent run took.

Activation of Low Memory Mode can indicate that the amount of memory allocated to the server is not sufficient to meet demands on the server. Iotellect Administrators may need to investigate how to reduce server load or increase available memory if Low Memory Mode is regularly active.

In some cases, such as when memory-intensive processes are running and variable values are being added to the cache faster than can be removed by the Memory Clearing operation, Iotellect activates Low Memory Mode, aimed at preventing Out Of Memory errors. Low Memory Mode reduces the memory usage of the server by halting client connections service, preventing new events from being submitted or processed, pausing updates from devices, and limiting some other server functions until enough memory is available for normal operation.

The duration and activation of Low Memory Mode is indicated in the Server Information variable Memory Clearing. Low Memory Mode activates as a protective measure to preserve minimal server operation.

Note that short Low Memory Mode periods lasting several seconds don’t necessarily point to abnormal server operation, but rather allow the server to handle high spikes in incoming data streams or internal activity peaks.

However, Iotellect administrators should investigate options for increasing available memory and reducing the memory load on the server if Low Memory Mode is activated for longer periods of time.

Was this page helpful?

Tell us more

Send

Thank you!