REU Site: BigDataX: From Theory to Practice in Big Data Computing at Extreme Scales

该奖项在伊利诺伊理工学院和芝加哥大学建立了一个新的本科生研究经验（REU）网站。该网站名为BigDataX，专注于极端规模大数据计算理论和实践的本科生研究。BigDataX包括一个由8名本科生和4名导师组成的多元化团队，分布在这两个机构。学生将在大数据领域进行研究，以及如何解决与运行时系统和存储系统的设计，分析和实施相关的问题，以支持大数据应用程序。这项工作包括使极端规模计算更易于处理，触及高端计算和数据中心的每一个分支。这些进步将影响国家一级的科学发现和经济发展，并将加强广泛的研究活动，从而有效地获取、处理、存储和共享来自许多学科的宝贵科学数据。该奖项的主要重点是促进以数据为中心的科学和技术计算观点，在分布式系统理论和实践的交叉点。项目团队已经确定了来自天文学、生物信息学、医学成像等许多学科的各种数据密集型应用，这些应用展示了大数据应用的特征。这项工作的重点是设计，实现和优化的运行时系统，以支持多任务计算和高性能计算的并行编程系统。 工作重点是分布式调度、动态配置、改进的容错能力以及对异构计算的支持。为了更好地支持大数据应用程序，该团队正在探索分布式文件系统和各种关键组件的改进，例如元数据管理，I/O访问模式合并，分布式出处以及探索分布式存储系统的新接口。这项工作涉及真实的应用程序、真实的数据和真实的测试平台，从IIT和芝加哥大学的小型集群、阿贡国家实验室的超级计算机到亚马逊AWS云。

This award establishes a new Research Experiences for Undergraduates (REU) site at Illinois Institute of Technology and the University of Chicago. The site, which is named BigDataX, focuses on undergraduate research in both the theory and practice of big data computing at extreme scales. BigDataX includes a diverse group of 8 undergraduate students and 4 mentors spread out over the two institutions. Students will conduct research in the area of big data and how it will address issues related to the design, analysis, and implementation of run-time systems and storage systems to support big data applications. This work includes making extreme scale computing more tractable, touching every branch of computing in high-end computing and datacenters. These advancements will impact scientific discovery and economic development at the national level, and they will strengthen a wide range of research activities enabling efficient access, processing, storage, and sharing of valuable scientific data from many disciplines. The primary focus of this award is to promote a data-centric view of scientific and technical computing, at the intersection of distributed systems theory and practice. The project team has identified various data-intensive applications from many disciplines such as astronomy, bioinformatics, medical imaging, that demonstrate characteristics of big-data applications. This work focuses on the design, implementation, and optimization of runtime systems to support parallel programming systems for both Many-Task Computing and High-Performance Computing. The work centers on distributed scheduling, dynamic provisioning, improved fault tolerance, and support for heterogeneous computing. To better support big data applications, the team is exploring distributed file systems and improvements to a variety of critical components such as metadata management, I/O access pattern coalescing, distributed provenance as well as exploring novel interfaces into distributed storage systems. This work involves real applications, real data, and real testbeds ranging from small clusters at IIT and UChicago, supercomputers at Argonne National Laboratory, to the Amazon AWS cloud.