Đang chuẩn bị liên kết để tải về tài liệu:
Why Markets Could (But Don’t Currently) Solve Resource Allocation Problems in Systems

Using market mechanisms for resource allocation in dis- tributed systems is not a new idea, nor is it one that has caught on in practice or with a large body of com- puter science research. Yet, projects that use mar- kets for distributed resource allocation recur every few years [1, 2, 3], and a new generation of research is exploring market-based resource allocation mechanisms [4, 5, 6, 7, 8] for distributed environments such as Planet- lab, Netbed, and computational grids. This paper has three goals. The first goal is to ex- plore why markets can be appropriate to use for allo- cation, when simpler allocation mechanisms exist. The second goal is to. | Why Markets Could But Don t Currently Solve Resource Allocation Problems in Systems Jeffrey Shneidman1 Chaki Ng1 David C. Parkes1 Alvin AuYoung2 Alex C. Snoeren2 Amin Vahdat2 and Brent Chun3 1 Harvard University 2 University of California San Diego 3Intel Research Berkeley Abstract Using market mechanisms for resource allocation in distributed systems is not a new idea nor is it one that has caught on in practice or with a large body of computer science research. Yet projects that use markets for distributed resource allocation recur every few years 1 2 3 and a new generation of research is exploring market-based resource allocation mechanisms 4 5 6 7 8 for distributed environments such as Planetlab Netbed and computational grids. This paper has three goals. The first goal is to explore why markets can be appropriate to use for allocation when simpler allocation mechanisms exist. The second goal is to demonstrate why a new look at markets for allocation could be timely and not a re-hash of previous research. The third goal is to point out some of the thorny problems inherent in market deployment and to suggest action items both for market designers and for the greater research community. We are optimistic about the power of market design but we also believe that key challenges exist for a markets systems integration that must be overcome for market-based computer resource allocation systems to succeed. 1 Is there a Problem During the past decade we have witnessed the emergence of systems that are owned deployed and used by multiple self-interested stakeholders. Consider the differences between traditional distributed systems and current distributed environments such as Planetlab Netbed and computational grids. Current environments have the following properties Many resources many users and more complicated needs. Multiple self-interested parties can simulate-nously supply and consume sets of resources e.g. machine time bandwidth . Users can demand large sets of .