Forked execution in runtime clusters and clouds
Forked execution is enabled in runtime clouds by default, which means that each process executes in a separate JVM. You can optionally enable forked execution in a runtime cluster.
By default, a runtime cluster's nodes execute processes within the same JVM as the node itself. However, you can enable an optional mode called forked execution that causes a node to execute each process in a separate JVM, forked each time that another process executes.
Forked execution cannot be enabled if multiple basic runtimes are running on the same machine. Java 8 JDK is required for runtime clusters to use forked execution.
The benefits of using forked execution are:
-
Currently-executing processes use more of the total memory of the machine.
-
The node is more robust because catastrophic failures (such as a JVM crash or OutOfMemoryException error) in a given process execution do not affect the main node process or any other currently-executing processes.
The drawbacks to using forked execution are:
-
A slight delay in process execution, due to operation system process startup time.
-
More machine resources are used than a single node.
You can enable forked execution in a runtime cluster on the Properties panel in Runtime Management. You can set forked execution properties for a runtime cluster or private runtime cloud on the Forked Execution Properties panel in Runtime Management.
Connector browse operations run in forked executions in private runtime clouds and therefore can use user-installed JAR files, installed at the account-level, and execute browse operations in a separate JVM.
