【作者】 张智勇；

【导师】 陆枫； 周艳红；

【作者基本信息】 华中科技大学 ， 计算机应用技术， 2007， 硕士

【摘要】 自从全基因组测序成为可能以来,基因组结构注释(包括了解基因组DNA中的基因组成、结构及其调控元件)成为生物信息学研究的重要问题,由于基因组结构注释系统的计算量庞大,需要基于高性能计算资源进行基因组结构注释计算。目前生物信息学研究机构普遍采用的高性能计算系统主要存在以下两方面问题:搭建成本高昂,性价比差,对预算有限的实验室是沉重负担;技术标准不开放,导致软件管理困难,硬件扩展性不佳,后期维护成本高昂。针对以上问题做了如下工作:搭建了一个适合生物信息学应用的基于开源集群资源管理中间件OSCAR的集群系统。对目前的硬件技术进行分析,在此基础上根据需求对硬件选型,特别针对目前统一架构集群的不足提出PC/SMP服务器混合架构的方案。对集群相关的软件进行了实验和选型,重点是集群资源管理中间件,选型过程中做了大量实验来测试系统的兼容性。对选型后的硬件和软件进行了系统集成,并对系统并行计算环境进行了测试。在集成过程中,克服了Linux软件的包依赖性等技术困难。使用国际标准的Linpack对集群的浮点计算性能进行了测试,对结果进行了分析和总结,验证了集群的计算性能和可扩展性是否符合要求。开发了求DNA-C/G含量的PBS程序cgpbs,对生物信息学者在集群系统的并行计算环境下部署复杂的生物信息学软件进行了一定探索。更多还原

【Abstract】 Since the whole genome sequencing possible, the genomic structure Notes (including understanding of the genomic DNA of the gene composition, Structure and control components) as bioinformatics research important issues, As genome annotation system structure of the enormous amount of computation, needing high performance computing resources for genomic structure calculation of the Notes.Currently bioinformatics research institutions commonly used in the high-performance computing systems mainly the following two aspects : high cost structures, poor cost performance, the budget is limited laboratory is a heavy burden; while technical standards is not open, leading to software management difficulties, hardware scalability and high maintenance costs later.To solve the problems above, doing the following : erection of a suitable application of bioinformatics cluster-based revenue resources management middleware OSCAR cluster system. The current hardware technology, the basis of demand for hardware selection, Special unified view of the current structure of the cluster of insufficient PC / SMP server hybrid structure of the program. Clusters of related software selection and the experiment, focusing on the cluster resource management middleware, Selection process has done a lot of experiments to test the system compatibility. Selection of the hardware and software for system integration and system parallel computing environment for the test. In the process of integration, overcome a Linux software package dependence technical difficulties. Use of international standards on the LINPACK floating-point calculation of the cluster properties of the test, Results of the analysis and review to verify the cluster computing performance and scalability whether it meets the requirements. Developing a DNA-C / G content of PBS cgpbs procedure, helping bioinformatics scholars to deploy complex bioinformatics software for a certain exploration in the cluster system of parallel computing environment.更多还原