Abstract
Reconfiguration delay is a defining systems parameter of photonic switching. Yet, it is almost universally modeled as a single constant in network design and circuit scheduling algorithms, namely the optical switching time reported by the device. We argue that this abstraction is no longer sufficient. From an application’s perspective, reconfiguration begins when a new circuit is requested and ends only when communication resumes over the new path. The latency observed by the application spans an end-to-end critical path across multiple hardware and software components, not just the optical switch.
This paper asks a simple question: what is reconfiguration delay, really? We argue that it should be understood as an end-to-end systems property rather than a device property. To make this concrete, we dissect the critical path of a circuit transition, spanning host-to-controller communication, controller processing, electrical signaling, and optical switching. Our measurements show that optical switching itself accounts for only a small fraction of the end-to-end latency, even before considering transceiver recovery and receiver-side reactions. This perspective raises a fundamental systems question: if most of the end-to-end delay lies outside the switch fabric, what should future photonic switching systems optimize? We conclude by outlining a research agenda for reducing end-to-end reconfiguration delay across the hardware–software stack.
Cite this work
BibTeX entry
@inproceedings{photonicsreconfigurationhotnets26,
author = {Rao, Deeksha P and Addanki, Vamsi},
title = {What Is Reconfiguration Delay, Really? An End-to-End View of Photonic Switching},
year = {2026},
booktitle = {Proceedings of the 25th ACM Workshop on Hot Topics in Networks}
}