CC* INTEGRATION-SMALL: ADIABATIC MICROSERVICE LEVEL LOAD BALANCED FORWARDING ON PISA SWITCH FOR ACCELERATING URGENT PROCESSES IN SCIENCE DATA CENTER NETWORKS

CC* 集成小型：PISA 交换机上的绝热微服务级负载平衡转发，用于加速科学数据中心网络中的紧急进程

• 批准号: 2346729
• 负责人: Javed Khan
• 金额: $ 50万
• 依托单位: Kent State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2024
• 资助国家: 美国
• 起止时间: 2024-06-01 至 2026-05-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2346729&HistoricalAwards=false
• 关键词: CC; INTEGRATION; SMALL; ADIABATIC; MICROSERVICE

Campus science data centers (Sci. DC) increasingly need support for more varied classes of workloads, mainly because many are interactive. Sci. DCs are using virtualization for better just-in-time distributed resource provisioning. Also on the horizon are the microservices that can accelerate collaborative development at scale. Virtualization and service-based applications demand advanced functionalities, particularly load balance (LB), while the requesters and servers are dynamically moving between physical hosts at uncompromising line rates. Very recently, using programmable PISA hardware, this team was able to design a load-balancing switch, QLB, that can distribute and load-balance load groups with a general weighted-cost multi-path (WCMP) scheme while concurrently forwarding to the network’s data plane at the total line rate of the hardware. It also has today's production-scale path and application cardinality, overcoming limitations of the software-based LBs or early PISA realizations. The project aims to develop a prototype of this WCMP load balancing switch that is deployable at Sci. DC. In conjunction with smart NICs, this smart switch will be ready to offer load balancing from inside the network in multiple modalities. If successful, this technology will significantly remove the delay and scalability limitations associated with the current generation software-based load balancing in campus Sci. DCs. It will also provide early insight into security, configuration, and maintenance issues related to deploying and operating programmable switches. In line with the goals of the CC* program, this will directly contribute towards coordinated campus-level cyberinfrastructure improvements, innovation, integration, and engineering for science applications and distributed research projects. The project will provide novel exposure to programmable cyberinfrastructure - the first for Kent’s Campus IT teams and many other campuses. The project will also have a significant broader impact: it will engage undergraduate students in interdisciplinary directed research internships with STEM science teams, which is especially effective in retaining and attracting underrepresented and disadvantaged students. It will empower science projects at Kent State across disciplines.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

校园科学数据中心（Sci。DC）越来越需要支持更多不同的工作负载，这主要是因为许多工作是互动的。科学。 DC正在使用虚拟化来更好地分布式资源提供。同样，可以大规模加速协作发展的微服务。虚拟化和基于服务的应用程序需要高级功能，尤其是负载余额（LB），而请求者和服务器则以毫不妥协的线路速率在物理主机之间动态移动。最近，使用可编程的PISA硬件，该团队能够设计负载平衡的开关QLB，该开关可以通过一般加权成本多路径（WCMP）方案分配和负载负载组，同时以硬件的总线路速率同时转发到网络的数据平面。它还具有当今的生产规模的路径和应用基数，克服了基于软件的LB或早期PISA实现的局限性。该项目旨在开发该WCMP负载平衡开关的原型，该开关可在SCI上部署。 DC。与智能NIC一起，此智能开关将准备以多种方式从网络内部提供负载平衡。如果成功，该技术将大大消除与校园SCI中当前基于软件的负载平衡相关的延迟和可伸缩性限制。 DCS。它符合CC*计划的目标，这将直接有助于校园级的网络基础设施改进，创新，整合和工程学，用于科学应用和分布式研究项目。该项目将为可编程网络基础设施提供新颖的曝光率，这是肯特校园IT团队和许多其他校园的第一个。该项目还将产生更广泛的影响：它将吸引本科生与STEM科学团队进行跨学科的指导研究实习，这在保留和吸引缺乏代表和干扰的学生方面特别有效。它将在跨学科中赋予肯特州立大学的科学项目。该奖项反映了NSF的法定使命，并通过使用基金会的知识分子优点和更广泛的影响审查标准来评估，被认为是宝贵的支持。