SHF: Small: Cooperative Memory Expansion (COMEX) for Networked Computing Systems via Remote Direct Memory Access

具有大型工作集的遗留应用程序和新兴应用程序在商用服务器上执行时，由于依赖于缓慢的磁盘交换分区的频繁页面错误，可能会表现出较差的性能。本研究旨在建立网络计算系统节点间的海量内存，用于保存在大型工作集的应用程序运行过程中从任何参与节点的主内存中取出的页面，实现协同内存扩展（COMEX）。它对推进网络计算系统的技术理解和科学前沿有着巨大的希望。鉴于远程直接内存访问（RDMA）技术被操作系统和网络设备供应商广泛采用，COMEX采用该技术在节点之间的页面传输中实现超低延迟，其操作系统内核和CPU可以完全绕过。该项目涉及五项技术挑战，这些挑战的解决方案将共同构成COMEX处理器的基础，能够通过更好地利用整体系统DRAM内存来提高执行性能和系统吞吐量。该项目还将改善路易斯安那大学拉斐特校区计算机系统和分布式计算的研究和教育活动，其成果有助于整合研究和教育，以丰富教学、培训和学习经验，并培养对NSF使命、路易斯安那州和国家至关重要的高素质未来劳动力。

Legacy and emerging applications with large working sets and, when executed on commodity servers, may exhibit poor performance due to frequent page faults which rely on the slow disk swap partition. This research aims to establish immense memory collectively across nodes of a networked computing system for holding pages evicted from the main memory of any participating nodes during application runs with large working sets, realizing cooperative memory expansion (COMEX). It has a great promise to advance technical understanding and scientific frontiers of networked computing systems. Given its wide adoption by operating systems and networking gear vendors nowadays, the remote direct memory access (RDMA) technology is adopted by COMEX for ultra-low latencies in page transfer between nodes whose OS kernel and CPU can be totally bypassed. This project deals with five technical challenges and the solution approaches to those challenges together will constitute the basis of the COMEX Handler, able to enhance execution performance and system throughput via better utilizing overall system DRAM memory on-demand. The project will also improve the research and educational activities on computer systems and distributed computing on the University of Louisiana at Lafayette campus, with its outcomes helping to integrate research and education for enriched teaching, training, and learning experience and to educate quality future workforce critical to the NSF mission, the state of Louisiana and the nation.