What is it about?

Packet scheduling is the essential network function for ensuring that data packets reach their destinations within their designated delivery deadlines. We show that the inclusion of packet route information in the scheduling algorithm enables the enforcement of ultra-low bounds on the delivery latency of packets that traverse multiple network domains. Our solution boosts the reliability and efficiency of a broad variety of applications, such as industrial automation, supercomputing, AI training, data streaming, and network function virtualization.

Featured Image

Why is it important?

Our novel scheduling algorithm only needs to be deployed at the ingress edge of each network domain, while the interior nodes of the domains can keep operating with conventional strict-priority scheduling of per-class packet queues. This enables the enforcement of ultra-low latency guarantees without refreshing the entire hardware base of the network, as instead required by all prior solutions. The operation of the overall framework scales well with the size of the network and the dynamicity of the data flow population because the introduction of every new data flow only requires the reconfiguration of its respective ingress scheduler.

Perspectives

This publication is the result of a long personal journey that started in another century with my first research projects on network routing and packet scheduling. It has grown to maturity through the fundamental contributions of my coauthors and other former colleagues. Even though the solution that we describe is built out of many small details, I believe that we can accurately claim that simplicity is the defining character of its design.

Andrea Francini
Nokia Bell Labs

Read the Original

This page is a summary of: Routing-Aware Shaping for Feasible Multi-Domain Determinism, December 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3636534.3696727.
You can read the full text:

Read

Contributors

The following have contributed to this page