A novel thread scheduler design for polymorphic embedded systems

Abstract:

The ever-increasing complexity of current day embedded systems necessitates that these systems be adaptable and scalable to user demands. With the growing use of consumer electronic devices, embedded computing is steadily approaching the desktop computing trend. End users expect their consumer electronic devices to operate faster than before and offer support for a wide range of applications. In order to accommodate a broad range of user applications, the challenge is to come up with an efficient design for the embedded system scheduler. Hence the primary goal of the thesis is to design a thread scheduler for a polymorphic thread computing embedded system. This is the first ever novel attempt at designing a polymorphic thread scheduler as none of the existing or conventional schedulers have accounted for thread polymorphism. To summarize the thesis work, a dynamic thread scheduler for a Multiple Application, Multithreaded polymorphic system has been implemented with User satisfaction as its objective function. The sigmoid function helps to accurately model end user perception in an embedded system as opposed to the conventional systems where the objective is to maximize/minimize the performance metric such as performance, power, energy etc. The Polymorphic thread scheduler framework which operates in a dynamic environment with N multithreaded applications has been explained and evaluated. Randomly generated Application graphs are used to test the Polymorphic scheduler framework. The benefits obtained by using User Satisfaction as the objective function and the performance enhancements obtained using the novel thread scheduler are demonstrated clearly using the result graphs. The advantages of the proposed greedy thread scheduling algorithm are demonstrated by comparison against conventional thread scheduling approaches like First Come First Serve (FCFS) and priority scheduling schemes.