机制设计理论及其在计算机网络协议设计中的应用研究

Theory of Mechanism Design and its Application in the Field of Protocol Design of Computer Networks

计算机网络协议的设计一般假设参与者是完全服从的。对于域间路由、IP多播、P2P文件共享等问题,这个假设并不成立。这些问题中各参与者都是自治的主体,其行为是自利的,以追求自身的利益最大化为目标。这给网络协议的设计带来挑战。机制设计理论用于设计多主体之间的博弈规则,以获得期望的结果。该理论为计算机网络中出现的这类问题的协议设计提供了方向。本文首先介绍了机制设计的基本概念,并以路由为例说明了其在计算机网络中的具体应用。传统的机制设计理论是微观经济学和博弈论的分支,在具体应用到计算机网络中需要处理很多新的问题,例如计算复杂性、隐私、分布式计算等。文章对近年来该领域的研究成果做了总结,并指出了未来的研究方向。

Agents are generally supposed to be obedient in designing the protocols of computer networks. However this supposition does not come into existence in such problems as domain routing, IP Multicast, P2P file sharing etc. The agents in these scenes are autonomous; their actions are selfish and their expected objects are to maximize their own benefits. New challenges appear in designing the protocols of computer networks. The theory of mechanism design aims at designing game rules of multi-agents, and obtaining desired outcomes. This theory gives new direction for these problems. The paper firstly introduces basic concepts in the theory of mechanism design, and illustrates the application of this theory, for example network routing. Traditional theory of mechanism design is the branches of microeconomics and game theory. Many questions come forth when applying the protocol design of computer networks, such as computation complex, privacy, distributed computation and so on. The paper summarizes the research results recently, and points out the future research directions.