Currently we stick to the strategy of deploying the match maker on the client itself. Partially, it is due to it being tightly coupled to the life-cycle manager that needs to be operating on the client.
What we actually want is:
- Convert the life-cycle manager into a kettle app so it could be deployed independently from the flow manager.
- Define an API between the flow manager and the life-cycle manager.
- Once the above tasks are done, the flow manager can be deployed separately from the life-cycle manager and thus does not need to reside on the client.