MassWare Reconfigurator and System-level Reflection

 

In traditional reflective middleware, there is only one component chain (or graph) in each middleware agent. The reconfiguration process is then to modify the chain (or graph) structure. However, due to the operation suspension and synchronous synchronization during the reconfiguration process, both the overhead and delay are too large to be tolerable for real-time vehicle applications.

By contrast, MassWare supports multiple component chains as shown in the Figure, each of which is located in an actuator. Every actuator is hooked to an event sensor that is described in the same policy. There is one and only one actuator active at any time and only the component chain in the active actuator processes the application data. To reduce the resource consumption by multiple chains, each actuator only maintains a chain of references, which point to marchlet instances, and a customized parameter list for each reference. When contexts change and satisfy a new sensor, the sensor will notify the decision engine for reconfiguration by switching active and inactive actuators. That is, the current active actuator is deactivated by suspending its operations as it reaches a safe state, where the reconfiguration can be safely performed, storing its run-time status, and disconnecting the component chain; the target actuator is then activated by connecting its components, restoring the run-time status, and resuming its operations.

 

Copyright(c) 2007 Lehigh University. All rights reserved.
Web master: Shengpu Liu (shl204@lehigh.edu)