Allow the configuration framework to manage instances of AppParts (and local-only couterpart, probably module subpart). See ManagedServiceFactory, the idea is that there are dynamic parts that can be creared or removed; for example the uSpace Gateway can have multiple connections, all share a set of DEs, but each have their own values; new connection instances can be added, or removed.
For this the framework needs 2 new methods: createInstance, and removeInstance. Probably defined as an advanced configuration interface.
Additionally the Configuration Manager should be able to determine where the part is instantiated, e.g.: there can be 2 instances of the uSpace Gateway in the uSpace, but some connections are set to one, and the rest to the other. This may also be configurable i.e: the configurable entity should say for each "instanciable part" the distribution policy: single instance, replicated (up to n times), full redundant (in all nodes).
There should be different algorithms to decide this. e.g.: Maybe for single, and replicated modes, the interface should return a fitness value, the node to return the highest fitness value should be assigned the new instance.
Allow the configuration framework to manage instances of AppParts (and local-only couterpart, probably module subpart). See ManagedServiceFactory, the idea is that there are dynamic parts that can be creared or removed; for example the uSpace Gateway can have multiple connections, all share a set of DEs, but each have their own values; new connection instances can be added, or removed.
For this the framework needs 2 new methods: createInstance, and removeInstance. Probably defined as an advanced configuration interface.
Additionally the Configuration Manager should be able to determine where the part is instantiated, e.g.: there can be 2 instances of the uSpace Gateway in the uSpace, but some connections are set to one, and the rest to the other. This may also be configurable i.e: the configurable entity should say for each "instanciable part" the distribution policy: single instance, replicated (up to n times), full redundant (in all nodes).
There should be different algorithms to decide this. e.g.: Maybe for single, and replicated modes, the interface should return a fitness value, the node to return the highest fitness value should be assigned the new instance.