Real-time Adaptive Scheduling Tasks Qos Guarantees

With the popularity of Internet and wide range of distributed applications, the technique of distributed objects middleware established multiple specifications and standards to support the interoperability of applications which are independent of hardware platform and operating systems. These specifications and standards often conceal details to achieve performance guarantee under functional interfaces and unable to satisfy application's requirement of Quality of Service without distinguishing functional paths and QoS path. QuO framework supports applications'QoS guarantee in distributed real-time environment. However, the framework fails to support real-time adaptive scheduling applications with stringent QoS requirements and the change of clients'runtime dynamic QoS requirements.Based on research of application scenario and QoS requirements of performance-critical avionics systems and analysis of current QuO framework, especially its quality description language, the paper proposes an extension of QuO framework which supports runtime application QoS needs dynamically. Application developers use QoS contract to specify possible QoS levels and correspondent parameters. QoS table is used to specify applications'runtime QoS requirements and adaptive changes of QoS requirements while meeting changes of external and system conditions. Additionally, both the client and the server can obtain runtime QoS information from QoS table. QoSMgr is responsible for establishing communication and transferring information among client,server and adaptive scheduling module. QoSMgr updates QoS table according to the changes of external environment, system condition, executed QoS level. When (1) external condition changes; (2) system condition changes; (3) the incoming of new task, QoSMgr will be triggered to call adaptive scheduling module to re-evaluate and adjust QoS levels to achieve best performance of the system.We designed extended QoS contract, QoS table and QoSMgr to implement the prototype system. For avionics system, we completed contract description, QoS table description, interface description and description of adaptive behaviors to support its runtime QoS requirement. In the framework, real-time adaptive applications with stringent QoS needs and resource-oriented applications with only runtime static QoS requirement could both be satisfied. Currently, the implementation of extended QuO framework is still under development. We are working on modification of QuO code generator.