DRACON: QoS Management for Large-Scale Distributed Real-Time Databases

摘要

The demand for real-time data services is increasingin many large-scale distributed real-time applicationsincluding advanced traffic control, global environmentcontrol, and the nation-wide electric power grid control.However, providing quality-of-service (QoS) for data servicesin such large-scale and geographically distributed environmentis a challenging task. In particular, both unpredictablecommunicational delays and computational workloads oflarge-scale distributed systems can lead to large number ofdeadline misses. We have designed a distributed real-timedatabase architecture called DRACON (Decentralized dataReplication And CONtrol), which enables QoS guaranteesfor large-scale distributed real-time applications. DRACONcouples cluster-based replica-sharing and a decentralizedcontrol structure to address communication and computationalunpredictability, simultaneously. The cluster-basedreplica-sharing mechanism not only enables scalable andbounded-delay access to remote data with high probability,but also decouples clusters to have less interaction, allowinga decentralized, thus scalable, QoS control structure. Thesimulation study demonstrates that DRACON’s decentralizedQoS control structure combined with a decentralizedreplica-sharing structure provides robust and predictableQoS guarantees in a highly scalable manner.