UC Davis

Modern scientific workflows are data-intensive, geographically distributed, and time-sensitive, demanding reliable, low-latency data transfers across high-performance networks. While current solutions, such as Energy Sciences Network (ESnet)’s On-Demand Secure Circuits and Advance Reservation System (OSCARS), offer static bandwidth reservation, they lack the fine-grained, deterministic guarantees essential for emerging real-time applications. To address this gap, we explore the applicability of Cyclic Queuing and Forwarding (CQF), a deterministic scheduling mechanism traditionally designed for Local Area Networks (LANs), to the unique challenges of Wide Area Networks (WANs). We propose a novel architecture that integrates Cyclic Queuing and Forwarding (CQF) with multi-path routing to enhance reliability and evaluate it through a comparative simulation against OSCARS on a realistic ESnet topology, considering both time-triggered and event-triggered flows. Our results demonstrate that with proper parameter tuning, CQF can achieve a performance comparable to that of OSCARS in terms of latency, acceptance rate, and network utilization. As one of the first efforts to adapt CQF for WAN-scale scientific workflows, this study highlights its potential to enable deterministic communication for high-volume, long-distance scientific networks.