CNPq: Service Differentiation, Efficiency, and Scalability in Distributed E-Commerce Servers

点击翻译按钮获取中文摘要

The focus of this proposal is the design and implementation ofcluster-based electronic commerce servers (or simply distributede-commerce servers) that are efficient, scalable, and capable ofservice differentiation. To achieve these properties, our serverswill efficiently adapt their quality of service, will efficientlyhandle any type of incoming request according to the client's profile,will not involve any centralized resources, and will cache both staticand certain dynamic content. The key challenges are in determiningthe set of policies and mechanisms that should be used todifferentiate the services provided and to implement truly efficientand scalable policies and mechanisms for request processing. At theend of the project, we will demonstrate two distributed e-commerceservers (an on-line store and an auction server) running on a clusterwith 64 processors.The contributions of this project will be the following: (1) Acomplete and general characterization of e-commerce workloads; (2) Anunderstanding of the issues involved in service differentiation fordistributed e-commerce servers, including the use of optimizationsbased on the past history of actions taken by high-priority clients;(3) The design, implementation, and evaluation of mechanisms forshared state maintenance and transaction support in distributede-commerce servers; and (4) The study of techniques and mechanisms forimproving the efficiency and scalability of distributed e-commerceservers.We believe that the potential implications of our research andcontributions are far reaching. In fact, we think that the impact ofour research can be substantial, given how significantly variouscompanies have come to rely on their on-line networking operations.Furthermore, our research can be useful for operating systems anddistributed systems teaching purposes, as our software infrastructurewill provide a tool for the study of distributed e-commerce servers.