CSR -- PDOS: Methodologies for Customization of Distributed Algorithms and Middleware

Kansas State University, Manhattan KS

Investigators

Abstract

Abstract: Distributed applications often require the underlying distributed algorithms and middleware to be customized to meet stringent performance requirements. This is typically addressed by manually re-implementing portions of existing algorithms, which is tedious and error-prone. The goal of this project is to develop a comprehensive framework for automated customization of distributed algorithms and middleware. The work aims at developing technologies for various aspects of customization, which include (a) techniques to design distributed algorithms and middleware amenable to customization, (b) domain-specific languages for specifying rules to identify customization opportunities, (c) infrastructure to analyze the specifications to determine when customization rules can be applied, (d) tools to perform the code transformations necessary to perform the optimizations, and (e) extensive evaluation of the developed technologies on a number of applications. The expected project outcomes include tools for automatic customization of algorithms, a study of the extent to which generic distributed algorithms can be customized automatically and a characterization of the types of customizations applicable to different classes of algorithms. This research will ease the development of efficient distributed algorithms to meet the requirements of different applications. Research results from this project will be leveraged in developing new courses on component based design of distributed systems, and in advancing multidisciplinary activities in the area of distributed sensor systems. Technology transition activities will be pursued in collaboration with our industrial partners. Efforts will be made to involve undergraduates and students from underrepresented groups in the research and educational activities.