CSR: Small: Designing Mechanisms for Resource Allocation in Competitive Real-Time Open Environments

Wayne State University, Detroit MI

Investigators

Abstract

Recent trends in computing systems, such as cloud computing, virtual execution environments, and grid computing, have been away from centralized designs towards highly extensible, open, and distributed systems comprising multiple independently-developed and operated subsystems. Such systems present a fundamental design challenge for applications that require real-time guarantees, due to the reliance upon all subsystem owners and developers to truthfully express their resource requirements. Currently, no real-time open system environment framework prevents a user from obtaining an ``unfair'' allocation of system resources by simply misrepresenting their subsystem's resource requirements. A solution is needed, that formally guarantees an efficient/equitable resource allocation among subsystems. The approach taken here is to apply a novel combination of real-time scheduling and game theory (i.e., algorithmic mechanism design) techniques, to design truth-inducing allocation mechanisms for shared real-time processing platforms. These techniques are proven and evaluated by developing a middleware implementation and a grid simulation for a real-time setting. A trustworthy scheme for allocating time-critical resources among independently developed subsystems is a key to reducing development and operational costs, and increasing the reliability of the complex computing systems upon which our society and economy increasingly depend. Beyond direct benefits of the technology, this project furthers educational goals by actively involving undergraduate and underrepresented students in research.