User-specified adaptive scheduling in a streaming media network

Abstract

In disaster and combat situations, mobile cameras and other sensors transmit real-time data, used by many operators or analysis tools. Unfortunately, in the face of limited, unreliable resources, and varying demands, not all users may be able to get the fidelity they require. This paper describes MediaNet, a distributed stream processing system designed with the above scenarios in mind. Unlike past approaches, MediaNet's users can intuitively specify how the system should adapt based on their individual needs. MediaNet uses both local and online global resource scheduling to improve user performance and network utilization, and adapts without requiring underlying support for resource reservations. Performance experiments show that our scheduling algorithm is reasonably fast, and that user performance and network utilization are both significantly improved.