CRII: CI: Exploring Advanced Cyber-Infrastructure Co-Design for Big Data Analytics

数字数据的激增为科学、工程和工业的所有领域提供了新的机会。然而，不断增加的数据量和速率，以及延迟和能源方面的相关成本，迅速主导并限制了数据分析应用程序以有效和及时的方式利用这些数据的能力。联合设计过程使科学家能够对软件和硬件中丰富的设计空间进行推理，预计将有助于构建一类新的网络基础设施。然而，当前的大数据分析解决方案存在重大局限性，这使得这些解决方案不适用于下一代网络基础设施。因此，在大数据分析范围内探索关键的协同设计问题变得至关重要。此项研究的首要目标是了解大数据分析在具有新兴架构的系统上与数据放置、移动和处理相关的性能、电源行为和权衡，并开发能够从根本上支持对正在进行的网络基础设施进行大数据分析的模型。与现有的工作不同，这项研究工作的重点是在大规模科学应用的背景下开发的联合设计过程。主要研究活动包括构建评估不同类别大数据分析的框架，从性能和能力方面表征数据分析应用程序，以及开发构建模型以了解和探索设计空间的方法。随着数据分析应用程序在广泛的领域变得越来越重要，开发大规模和可持续的平台和软件基础设施以支持这些应用程序的能力具有巨大的潜力，可以推动科学和商业领域的研究和创新。本项目提供对大数据分析和未来网络基础设施设计的硬件/软件特征和要求的了解。这些研究活动与研究生和本科生的研究相结合，并利用罗格斯大学的少数族裔学生外联网络。

The proliferation of digital data provides new opportunities in all areas of science, engineering and industry. However, the increasing data volumes and rates, and the associated costs in terms of latencies and energy, quickly dominate and limit data analytics applications' ability to leverage this data in an effective and timely manner. The co-design process enables scientists to reason about the rich design spaces available in software and hardware, and is expected to help in constructing a new class of cyber-infrastructure. Nevertheless, current solutions for big data analytics have important limitations that make these solutions infeasible for the next generation of cyber-infrastructure. As a result, exploring key co-design issues in the scope of big data analytics has become critical.The overarching goals of this research are to understand the performance and power behaviors and tradeoffs related to data placement, movement, and processing of big data analytics on systems with emerging architectures, and to develop models that can fundamentally enable big data analytics on ongoing cyber-infrastructure. In contrast to existing work, this research effort focuses on the co-design process that has been exploited in the context of large-scale scientific applications. Key research activities include building a framework to evaluate different classes of big data analytics, characterizing data analytics applications in terms of performance and power, and developing a methodology to construct models to understand and explore the design space.As data analytics applications become increasingly important in a wide range of domains, the ability to develop large-scale and sustainable platforms and software infrastructure to support these applications has significant potential to drive research and innovation in science and business domains. This project provides an understanding of hardware/software characteristics and requirements for big data analytics and future cyber-infrastructure design. These research activities are integrated with graduate and undergraduate student research, and leverage minority student outreach networks at Rutgers.