Dealing with Memory Usage
From point of view of an underlying operating system Iotellect Server, as any other Java software, consumes as much memory as it's allowed to. Never much more, less in rare cases. The maximum allowed amount is specified by the -Xmx setting defined in Launch Properties File.
In most cases, the Iotellect installer sets this value to 50% of RAM available on the machine. Thus, if a system administrator will check server memory usage by using Windows Task Manager or Linux top command he will likely see that Iotellect Server's memory footprint is quite high.
However, this has nothing to do with real memory usage! Java Virtual Machine (JVM) running Iotellect Server has its own dynamic memory management system. The system is quite complicated, but, in very simple words, it has a periodic process called Garbage Collection. This process cleans up all temporary memory allocations, allowing us to see how much memory is, indeed, used by the server on a long-term basis.
The only way to figure out the real server memory usage is a sequence of two operations:
Execute Run Garbage Collection action from Root context to force full JVM garbage collection cycle
Check the value of Memory Usage, % of Maximum Allowed field in the Server Status variable of the Root context, e.g. by executing View Server Information action from the Root context
 | Checking the Memory Usage parameter without prior run of the garbage collection cycle will show random values ranging from the actual memory usage to 100%. This result isn't useful for any kind of interpretation. |
If the actual server memory usage (observed just after a garbage collection cycle) is above 90%, the server's JVM suffers from a memory shortage, effectively spending more time on the garbage collection process and raising CPU load.
If a JVM can not clean up enough memory to proceed with normal server operation, it will consume up to 100% of CPU power for the garbage collection process. Due to JVM’s memory management nature, this may happen before even the very first memory shortage error will be logged!
Thus, if actual memory usage raises over 90% of the allowed maximum, the value of -Xmx should be increased to allow effective server operation. If the new -Xmx value is over 70-80% of available RAM size, server RAM should be expanded to leave enough memory for Iotellect Server, operating system, and third-party software.
Vice versa, if the actual memory usage percent is quite low and the server is already running in production mode, the value of the -Xmx parameter can be decreased to free more memory (e.g. for server database or other applications). For example, if memory usage is 30%, the value of -Xmx can be divided by two, and the lowest memory usage will raise to 60% of -Xmx, which is still safe.
Per-Context Memory Usage
It's also possible to analyze the approximate size of JVM heap memory retained by each individual context.
Per-context memory usage is displayed in Retained Memory column of the statistical table accessed via the View Context Statistics action of Root context.
Was this page helpful?