XD Metrics Service (XMS)

XD度量服务（XMS）是技术审计服务（TAS）的更新项目，旨在提高NSF XD计算资源网络的运行效率和管理。XMS建立并扩展了TAS项目的成功，例如XDMoD工具的开发。该工具为XD及其最大的项目XSEDE的涉众提供了访问有关XD资源和XSEDE相关服务的利用率、性能和服务质量的数据的便利。虽然最初的项目重点是XD社区，但持续的努力意识到，这样的资源管理工具对于高性能计算中心以及管理复杂IT基础设施的其他数据中心也非常有用。为了抓住这个机会，正在开发Open XDMoD，这是该工具的开源版本。Open XDMoD已经被许多学术和工业HPC中心使用。XMS项目对XDMoD进行了扩展，超出了其最初的目标，从而增加了XD的实用性，并使其成为网络基础设施的综合资源管理工具。一个例子是通过“TACC_Stats”将作业级性能数据合并到XDMoD中。该功能使XDMoD能够识别性能较差的应用程序、提高吞吐量、描述系统工作负载的特征，并提供对未来资源获取规范至关重要的指标。考虑到当今高性能计算系统的规模，即使是吞吐量的适度增加也会对科学和工程研究产生重大影响。例如，对于XD网络，系统性能每提高1%就意味着可以额外分配1500万CPU小时的计算机时间用于研究。

The XD Metrics Service (XMS) is a renewed project of the Technology Audit Service (TAS), which aims at improving the operational efficiency and management of NSF's XD network of computational resources. XMS builds on and expands the successes of the TAS project, such as the development of the XDMoD tool. This tool provides stakeholders of XD and its largest project, XSEDE, with ready access to data about utilization, performance, and quality of service for XD resources and XSEDE-related services. While the initial project focus was the XD community, the ongoing effort realized that such a resource management tool would also be of great utility to high performance computing centers in general, as well as to other data centers managing complex IT infrastructure. To pursue this opportunity, Open XDMoD was being developed, which is an open source version of the tool. Open XDMoD is already in use by numerous academic and industrial HPC centers. The XMS project expands XDMoD beyond its original goals, so as to increase its utility to XD and move it into the realm of a comprehensive resource management tool for cyberinfrastructure. One example is the incorporation of job-level performance data through "TACC_Stats" into XDMoD. This functionality provides XDMoD with the ability to identify poorly performing applications, improve throughput, characterize the system's workload, and provide metrics critical for the specification of future resource acquisitions. Given the scale of today's HPC systems, even modest increases in throughput can have a substantial impact on science and engineering research. For example, with respect to the XD network, every 1% increase in system performance translates into an additional 15 million CPU hours of computer time that can be allocated for research.