Abstract: Commercialization of large scale grid systems requires the provision of service composition to achieve agreements on end-to-end QoS demands for the service consumer and the service providers.Efficient negotiation and guarantee mechanisms for service composition are playing an increasingly important role in the performance of the system.This paper proposes a grid contractual computing model（GCCM） to automatically and dynamically achieve agreements on quality of service constraints for individual services in a service composition.The GCCM innovatively exploits the agent technology to fulfill autonomous grid contractual computing negotiation，which follows a progressive FIPA protocol with two-phase commits.The innovative service contracts introduce the notion of soft service contracts（SSC） and hard service contracts（HSC）.The soft contract generated before dynamic binging is a placeholder，and only that through the contract commit mechanism does the contract harden into an actual promise of services.

Key words: grid contract, contractual computing, agent technology, service composition, dynamic negotiation

摘要： 随着网格商业化的不断深入，网格计算要求：服务提供者和消费者间在组合服务中需达成端对端的高服务质量协议，高可靠性的组合服务协商机制和保障机制在网格计算中扮演着越来越重要的角色。所提出的网格合同计算模型，在组合服务过程中提供统一的用户接口，使交易双方自动动态协商满足每个单独服务的服务质量约束条件。在网格合同计算模型中，利用代理技术实现了自治网格合同计算协商机制，该协商机制遵循改进的两阶段提交FIPA协议。在服务合同中导入了软服务合同SSC和硬服务合同HSC概念，并指出动态绑定服务之前所生成的软服务合同SSC仅仅起到占位符的作用，只有在绑定时通过“硬化”才获得真实的服务承诺。

关键词: 网格合同, 合同计算, 代理技术, 服务组合, 动态协商