CSR: Small: IO-Efficient Computer System for Graph Analytics

CSR：小型：用于图形分析的 IO 高效计算机系统

• 批准号: 1717774
• 负责人: H. Howie Huang
• 金额: $ 35万
• 依托单位: George Washington University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-01 至 2022-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1717774&HistoricalAwards=false
• 关键词: CSR; Small; IO; Efficient; Computer

As graphs are widely used to represent entities and relationships in information, social, biological, and cyber-physical systems, knowledge generated from graph analytics is crucial to the success of sciences, commerce, health care, and national security. In graph analytics systems, input/output (IO) remains the main research challenge because graph processing requires a relatively small amount of CPU cycles. This research investigates multiple facets of the IO challenge from cache, memory to disks. The proposed research delivers IO-efficient computer systems for graph analytics, advancing the state of the art in this field. The research produces a comprehensive performance analysis of graph analytics systems. A novel IO architecture for graph data and metadata is developed to efficiently utilize IO bandwidth of underlying storage devices and maximize the hit ratios in memory and caches. New designs of graph systems deliver transformative innovations in the area of big data processing. During the course of this project, undergraduate and graduate students, especially under-represented students, participate in computer systems and big data research. In addition, the PI mentors students in local high schools and collaborates with industry.

由于图形被广泛用于表示信息，社会，生物和网络物理系统中的实体和关系，因此从图形分析中生成的知识对于科学，商业，医疗保健和国家安全的成功至关重要。在图形分析系统中，输入/输出（IO）仍然是主要的研究挑战，因为图形处理需要相对少量的CPU周期。本研究调查了IO挑战的多个方面，从缓存、内存到磁盘。拟议的研究为图形分析提供了IO高效的计算机系统，推进了该领域的最新技术水平。这项研究产生了一个全面的性能分析图分析系统。为了有效利用底层存储设备的IO带宽，最大化内存和缓存中的命中率，提出了一种新的图形数据和元数据IO架构。图形系统的新设计在大数据处理领域提供了变革性的创新。在这个项目的过程中，本科生和研究生，特别是代表性不足的学生，参与计算机系统和大数据研究。此外，PI还指导当地高中的学生，并与行业合作。

项目成果

Aquila: Adaptive Parallel Computation of Graph Connectivity Queries

• DOI: 10.1145/3369583.3392690
• 发表时间: 2020-06
• 期刊: Proceedings of the 29th International Symposium on High-Performance Parallel and Distributed Computing
• 作者: Yuede Ji;Howie Huang

Mnemonic: A Parallel Subgraph Matching System for Streaming Graphs

助记符：流图的并行子图匹配系统

• DOI: 10.1109/ipdps53621.2022.00038
• 发表时间: 2022
• 期刊: 2022 IEEE International Parallel and Distributed Processing Symposium (IPDPS
• 作者: Bhattarai, Bibek;Huang, Howie
• 通讯作者: Huang, Howie

Faster Stochastic Block Partition Using Aggressive Initial Merging, Compressed Representation, and Parallelism Control

使用积极的初始合并、压缩表示和并行控制实现更快的随机块分区

• DOI: 10.1109/hpec49654.2021.9622836
• 发表时间: 2021
• 期刊: 2021 IEEE High Performance Extreme Computing Conference (HPEC
• 作者: Uppal, Ahsen J.;Choi, Jaeseok;Rolinger, Thomas B.;Howie Huang, H.
• 通讯作者: Howie Huang, H.

GraphOne: A Data Store for Real-time Analytics on Evolving Graphs

GraphOne：用于实时分析演化图的数据存储

• 发表时间: 2019
• 期刊: Proceedings of the 17th USENIX Conference on File and Storage Technologies
• 作者: Kumar, Pradeep;Huang, H. Howie
• 通讯作者: Huang, H. Howie