【作者】 刘文懋；

【导师】 方滨兴；

【作者基本信息】 哈尔滨工业大学 ， 计算机应用技术， 2008， 硕士

【摘要】 生物信息学已成为能够改变科学发展的决定性力量之一,网格计算为生物信息应用提供了强大的计算和存储平台。然而,网格应用首先需要解决服务集成、资源异构、作业管理和调度等问题,同时向生物学家屏蔽上述复杂的底层实现。一个可行的方法是建立一个Web门户作为中间媒介。本文介绍了一个在网格环境下的生物门户站点BioLab项目,系统定义了服务的提供者、部署者和使用者三种用户角色;设计了网格环境下的服务和资源整合机制,实现了用户管理以及作业调度控制等功能;同时,系统提供了标准Web Service的接口,增加扩展性。我们分析了启发式调度算法,选择在线调度而非批处理调度以提高系统响应速度。实验表明,在较大的作业平均到达时间条件下,非阻塞的调度方式的作业总完成时间比阻塞的调度方式小;当作业平均到达时间增加时,作业总运行时间也会增加,但速度会变缓;在较小的作业平均到达时间条件下,各种在线调度启发式算法差别不大,而在较大的作业平均到达时间条件下,随机性好的随机调度算法和最大可用内存容量优先调度算法有更小的作业总完成时间。更多还原

【Abstract】 Grid provides powerful computing and storage platform for Bioinformatics, which has become one of the leading forces changing the way in which science is now conducted. However, grid appliances first need to solve problems such as service integration, heterogeneous resource, job management and scheduling while masking complex implements, where a web portal becomes a feasible solution.We introduce BioLab, a grid bioinformatics portal. We define three roles in the system: service providers, service distributers and service users, design integrated mechanism of services and resources and implement modules such as user management and job scheduling and job management. We also provide a non GT Web Service API.In the experiments, we analyze heuristic scheduling algorithms and choose online scheduling algorithms rather than batch scheduling for response speed. We find that in large average job arrival time, unblocked scheduling is better than blocked scheduling, as the average job arrival time increases, Makespan of workflow also increases but the trend becomes slower, in small average job arrival time, all online scheduling algorithms are nearly the same while in large average job arrival time, random algorithm and max available memory scheduling algorithm greatly outperforms the other.更多还原