150 likes | 273 Views
Endowing PauWare Components with Autonomic Capabilities. Cyril Ballagny , Nabil Hameurlain, Franck Barbier LIUPPA, France M-ADAPT’07 (ECOOP), Berlin, 30 th July 2007. The PauWare component model. A PauWare component encapsulates a statemachine to realize its behavior
E N D
Endowing PauWare Components with Autonomic Capabilities Cyril Ballagny, Nabil Hameurlain, Franck Barbier LIUPPA, France M-ADAPT’07 (ECOOP), Berlin, 30th July 2007
The PauWare component model • A PauWare component encapsulates a statemachine to realize its behavior • The provided services are mapped to statemachine events • Communication between components relies on event broadcasting • The statemachine model persists at runtime : enable to observe/control the component state
Our vision • In a self-managing context, we need a way to • Locally control adaptation on anormal events • Globally insure the coherence of adaptation and reconfiguration of a component assembly • Adding a feedback control loop to each PauWare component • Making PauWare components communicate respecting an interaction protocol : a set of rules to be respected by individual entities using structured messages. • enable validation and verification of properties such as termination
PauWare component control loop : UML Metamodel • A set of sensors to detect anomalies • An aggregator to centralize information coming from sensors • An evaluator to apply a policy and choose a corrective action • A set of effectors to realize the plan decided by the evaluator
Communication between autonomic PauWare components Two PauWare components
Communication between autonomic PauWare components Feedback loop elements
Communication between autonomic PauWare components FIPA request messages
Communication between autonomic PauWare components FIPA request messages
Communication between autonomic PauWare components FIPA request messages
Communication between autonomic PauWare components FIPA request messages
Communication between autonomic PauWare components FIPA request messages
Conclusion • We use state machine based components and executable models as a starting point • We endow each PauWare component with a feedback loop • We enrich communication between PauWare components with MAS interaction protocol
On the use of agent interaction protocols :FIPA Request Interaction Protocol • Agent protocols are well formalized and enable negociation, coordination, agreement … • Consider autonomy : Refuse vs Agree • Consider fallibility Failure