Model-driven QoS Provisioning for Distributed Real-time and Embedded Systems

Real-time and Embedded Systems JAIGANESH BALASUBRAMANIAN, SUMANT TAMBE, ANIRUDDHA GOKHALE, DOUGLAS C. SCHMIDT Department of EECS, Vanderbilt University, Nashville, USA and BALAKRISHNAN DASARATHY, SHRIRANG GADGIL Tel ordia Te hnologies, Pis ataway, NJ, USA Distributed real-time and embedded (DRE) systems onsist of performan e-sensitive appli ations that are deployed in resour eonstrained environments and have diverse CPU and network quality-of-servi e (QoS) requirements. Coordinated allo ation of CPU and network resour es are required so that multiple DRE appli ations an e e tively share the available resour es and have their QoS requirements satis ed end-to-end. Although CPU QoS me hanisms, su h as bin-pa king algorithms, and network QoS me hanisms, su h as di erentiated servi es (Di Serv), an manage a single resour e in isolation, relatively little work has been done on QoS-aware me hanisms for managing multiple heterogeneous resour es in a oordinated, integrated, and non-invasive manner to support end-to-end appli ation QoS requirements. This paper provides two ontributions to the study of middleware that supports QoS-aware deployment and on guration of appli ations in DRE systems. First, we present a model-driven omponent middleware framework alled NetQoPE and des ribe how it shields appli ations from the omplexities of lower-level CPU and network QoS me hanisms by simplifying (1) the spe i ation of per-appli ation CPU and perow network QoS requirements, (2) resour e allo ation and validation de isions (su h as admission ontrol), and (3) the enfor ement of perow network QoS at runtime. Se ond, we empiri ally evaluate how NetQoPE provides QoS assuran e for appli ations in DRE systems. Our results demonstrate that NetQoPE provides exible and non-invasive QoS on guration and provisioning apabilities by leveraging CPU and network QoS me hanisms without modifying appli ation sour e ode.