Process model runtime configuration with no commits to disk

Is it possible to configure the process run-time environment to work without causing writes to disk (inter step transitions) and commits to database (process audit log operations, inserting step input pipeline to database)?

What I’d like to achive in this way is to reuse the join feature of the models and correlation services to implement waiting for partial responses from various systems that I need to wait for in order to create a final response to a business query.
I can model such a process currently and use it but due to many disk writes that are performed during at process run-time I cannot meet the time constratints related with the final response.