In-Memory Workflows - How it Works
When workflows are set to in-memory, all the workflow data is retrieved from the cache and you can run workflows without persisting workflow state information.
The following object types are put in memory:
Workflow requests that are written and read from a file are now part of the JMS payload. Input and output parameters that are XML files, are stored in memory and are not written to a disk. When a workflow is executed, the workflow states and documents are managed in the local cache.
You can enable or disable the execution of in-memory workflows (in which case, workflows are executed in the transactional mode) using Configurator. For more information, see Configuring In-memory Execution through the Configurator.
After a message is received, and the associated event and document created and stored in the distributed cache, if the workflow execution mode is set to in-memory, the event and document are copied from the distributed cache to the local cache. Activities in the process are executed sequentially in a single thread on a node in the cluster.