Implement a deployment model where the TarMK can run in a dedicated JVM on its own box allowing higher layers to connect to it through a "database like connection".
This should start out as a research effort allowing us to experiment with alternative deployment models, concurrent connections and possibly clustering of the TarMK. We would be mainly interested in:
- Independent scalability of the TarMK
- (Performance) impact of the remoting
- Serialisability of transactions (from different connections)
- Ideas and ways to implement clustering for the TarMK