CSR: Medium: MAD Systems

点击翻译按钮获取中文摘要

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).We increasingly rely on MAD Services, systems that span Multiple Administrative Domains. Examples include cloud computing, peer-to-peer streaming, Internet routing, and wireless mesh routing. Unfortunately, when different parties control a system's nodes, the nodes may undermine the desired service either by malfunctioning or by selfishly maximizing their own benefits to the detriment of the broader system (e.g., a tragedy of the commons.)The central thesis of this research is that any approach to building robust MAD services must account for both selfish behaviors and Byzantine malfunctions.This research therefore brings together ideas from game theory and Byzantine fault tolerance to (1) develop the new theoretical framework for MAD distributed systems; (2) design techniques that rigorously address key practical issues in building MAD systems, such as heterogeneity, locality, and node churn; and (3) construct significant prototypes of dependable MAD systems that tolerate selfish and Byzantine behaviors and whose costs and performance are comparable to less robust systems deployed today.This work seeks to have broader societal impact in at least two dimensions. First, it will develop the theory and practice of MAD systems at a time when crucial personal, business, and societal needs are increasingly met by MAD services. Second, we expect a significant educational impact from this work, as graduate and undergraduate students will participate in a research plan that will enhance skills in both practice and theory.