Event-based QoS for a distributed continual query system

Continual queries for information flows are attracting increased interest. However, two important characteristics of information flows are not adequately addressed by current continual query packages: distribution and quality-of-service (QoS). In this paper, we use Infopipes, an abstraction for information flows, and the Infopipes stub generator to create a distributed version of the linear road benchmarking application around an existing continual query server (CQ server). Then, we install a QoS event generator/monitor pair that observes system response latencies. Receipt of each query result generates a feedback event reporting the latency (response time) for that request. If latencies exceed a threshold as parameterized by a WSLA, then the Infopipes controlling the CQ server can attempt to restore latencies to an acceptable level by generating an adaptation event for the Infopipes managing the CQ server.