AdaptStream: towards achieving fluidity in adaptive stream-based systems

Abstract

Stream-based systems are frequently subject to changes in their operational environments due to fluctuations in the available computation and communication resources. Dynamic adaptation is a mechanism to improve the tness of such systems. However, adaptation can block one or more streams thus inadvertently a ecting the timeliness properties of streams. This paper describes AdaptStream, an adap- tation framework that provides timeliness support for stream-based adaptations. We introduce the concept of fluidity to measure the temporal alignment of stream synchronization during adaptation. We present a scheduling algorithm that calculates the time-bounded schedule of adaptation actions on multiple streams to achieve the fluidity requirement that is traded o against available resources and the smoothness requirement of individual streams.