AmLight-INT: In-band Network Telemetry: To enhance the SDN network telemetry capability of the AmLight Express and Protect (AmLight-ExP) project with in-band Operation/Administration/Management (OAM) to support multiple 100G real-time big-data transfers with complex SLA requirements between the U.S. and South America
Intellectual Merit: With current technology, based on legacy systems and monitoring protocols, it is very complex and time-consuming to troubleshoot network transient events. These short-term and sporadic degradations in network performance often go undetected and have a high impact (packet loss) over big-data transfers in networks with a high delay-bandwidth product, such as AmLight-ExP’s international links. AmLight-INT proposes to respond to these challenges by changing the forwarding pipeline to include telemetry functions directly on the data plane. Generating and exporting telemetry metadata directly on the data plane can be performed at line-rate without affecting at all the control plane functions. Gathering telemetry metadata directly from the data plane will allow the AmLight-ExP SDN network to receive and process telemetry information in real-time with the capacity of feeding the AmLight-ExP SDN controller and user applications with the network status for fast reaction. Applications, such as Large Survey Synoptic Telescope, would benefit from sub-second network status to compensate interface queue utilization, packet drops, and link utilization. With real-time network utilization, the AmLight-ExP SDN controller will be able to optimize interface buffer sizes to accommodate multiple high-priority flows in parallel, avoiding tail drops that ultimately lead to poor data transfer performance.
Scope: AmLight-INT will be a coordinated project, using NSF investment, with industrial partners to develop a hardware-software solution capable of performing in-band network telemetry using the P4 language. NSF support will be used to acquire P4-capable switches, connectivity components, and staff support for research and software development during the twenty-four-month project.
Broader impacts: By determining how to deploy and use in-band network telemetry, AmLight-INT will bring a new resource to science and education and will transfer the understanding of this new technology to the R&E network community. As vendors and network operators do not deploy data plane-based in-band network telemetry, this project will be an opportunity for U.S. Computer Science and Engineering graduate students to be in on the ground floor of international technology innovation. The result will be enabling network monitoring and troubleshooting to an unprecedented level. We expect that at the end of this project, a new network design paradigm will have been created, presented, and used by network operators. The results will further benefit the future needs for high-speed international connectivity by creating an approach that efficiently mitigates transient network issues. Moreover, it will underscore that research partnership are possible between switch manufacturers, operators, and the research community. As the AmLight-INT will be exclusively based on standard protocols/open source solutions, other network manufacturers will profoundly benefit from such project, enhancing the whole networking ecosystem.