Proteins to Population:A metacluster for Bioinformatics

描述(申请人提供)：这项提案寻求资金，以建立一个“元集群”，以支持犹他大学正在开发的新一代生物信息学应用程序。元集群超越了单一通用计算集群的概念，专门针对针对其进行优化的问题定制资源和计算元素。 这些资源包括一个计算“循环农场”、一个具有快速节点间通信以优化并行处理的紧耦合集群、一个紧密耦合的数据密集型存储/处理单元和两个大型存储设施共同构成的元集群。在构建这样一个集群的同时，元集群环境的一个关键组成部分是促进和管理研究人员对这些资源的访问。使用犹他州高性能计算中心(CHPC)正在开发的高度先进的调度技术(MAUI/Silver)，每个单独的元素都将无缝地集成(从用户的角度)到似乎是单一资源的地方。基于我们在构建和操作商用硬件群集方面的丰富经验，我们设计的元群集不仅为我们的新兴应用程序集提供了必要的资源，而且以最具成本效益的方式实现了这一点。拟议的生物信息学应用程序表现出很大程度的可变性，具有不同的计算、可视化、数据和并行处理需求，因此需要访问相当不同的资源才能以最佳方式运行。这些应用范围从人群的流行病学研究到膝部力学的研究或器官水平系统的研究，再到各种生物分子系统的详细原子模拟。

DESCRIPTION (provided by applicant): This proposal seeks funding to build a "meta-cluster" to enable the new generation of bioinformatics applications under development at the University of Utah. Meta-clustering moves beyond the concept of a single general purpose computational cluster to specifically tailor resources and computational elements towards problems for which they are optimized. These resources will include a computational "cycle farm", a tightly coupled cluster with fast internode communication for optimitized parallel processing, a visualization cluster a tightly coupled data intensive storage/processing unit and two large storage facilities together forming the meta-cluster.Beyond building such a cluster, a key component of the meta-cluster environment is facilitating and managing access to these resources by individual researchers. Using highly advanced scheduling techniques (Maui/Silver) under development at the Center for High Performance Computing (CHPC) at Utah, each of the separate elements will be seamlessly integrated (from a user's perspective) into what appears to be a single resource. Based on our extensive experience in building and operating clusters of commodity hardware, we have designed the meta-cluster to not only provide the necessary resources for our emerging set of applications but done so in the most cost effective manner. The proposed bio-informatic applications exhibit a large degree of variability, have varied computational, visualization, data and parallel processing needs, and therefore require access to quite heterogeneous resources to perform optimally. These applications range from epidemiological studies of populations to investigations of the mechanics of knees or studies of organ level systems down to detailed atomistic simulations of various biomolecular systems.