Adaptive resource utilization via feedback control for streaming applications

A large emerging class of interactive multimedia streaming applications that are highly parallel can be represented as a coarse-grain, pipelined, data-flow graph. One common characteristic of these applications is their use of current data: a task would obtain the latest data from preceding stages, skipping over older data items if necessary to perform its computation. When parallelized, such applications waste resources because they process and keep data in memory that is eventually dropped from the application pipeline. To overcome this problem, we have designed and implemented an adaptive resource utilization (ARU) mechanism that uses feedback to dynamically adjusts the resources each task running thread utilizes so as to minimize wasted resource use by the entire application. A color-based people tracker application is used to explore the performance benefits of the proposed mechanism. We show that ARU reduces the application's memory footprint by two-thirds compared to our previously published results, while also improving latency and throughput of the application.