Our application needs to process incoming XMLs, and using XML-to-domain mapping is our natural choice rather than doing XML handling ourselves. Great time saver! But over time, we observe that the performance takes a huge hit because everytime an XML is received (which is 100,000 times a day), Mendix creates a FileDocument, writes it to disk. This gets terribly slow over time, as the server has 100s of nested directories. Generating the auto-number for each file is also intensive operation. Further, those XMLs aren't of any use. Domain (database) is what we care about. Of course, we can delete them after the microflow has finished working. So, the concern isn't about server disk space, but improving the speed of this microflow. Is there a way to achieve XML to Domain mapping without using any FileDocument?
Based on your comments, you could redesign your process. My approach would be:
You could publish a webservice in Mendix and call this one from your JMS connector. See documentation how to publish a webservice with a microflow.
Basically you need to model all entities in a domain model and select the main entity in your published webservice microflow. Then you can configure which attributes/associations are available in the webservice. You need to map the temporary objects in your microflow to real objects in your domain model. After that, call the webservice in your jms connection.