Hierarchical QoS integration for real-time systems

Providing QoS (Quality of Service) in real-time systems is crucial in the environment where applications are time sensitive; system resources (CPU, network bandwidth, storage etc.) are diverse, dynamic and related to each other.; In this research, we perform QoS integration along two dimensions and in a hierarchical way. The first integration combines different scheduling paradigms using a sliding-window approach. A new Priority-based Weighted Fair Queueing (PWFQ) is proposed to demonstrate the way how the integration is done by combining priority-driven and share-driven scheduling under the context of network packet scheduling. PWFQ is generalized to a hierarchical architecture through which scheduling and budget group are proposed to enable hierarchical scheduling and resource sharing. Another dimension we study integrates multiple types of resources including both CPU and network. We extend a two-level scheduling hierarchy to a general resource management framework which includes one resource allocator and multiple resource dispatchers. In the framework, multiple resources are managed together using uniform APIs. The relationship between multiple resources can also be defined. Results show that hierarchical integration in both dimensions provides effective solutions to support QoS in real-time systems.