Abstract
IEEE International Conference on High Performance Switching and Routing 2023 was organized in Albuquerque, NM, USA, June 5–7, 2023, and was the 24th edition of the conference. More than 47 participants attended the conference in person and several more participants joined remotely. The Organizing Committee consisted of Dr. Dimitrios Koutsonikolas, Dr. Carla Fabiana Chiasserini, and Dr. Gang Li TPC Co-Chairs, Dr. Anna Maria Vegni and Dr. Aris Leivadeas, Workshop Co-Chairs, Dr. Chiara Boldrini, Tutorial Chair, Dr. Xiang Sun, Posters and Demos Chair, Dr. Pantelis Frangoudis, and Dr. Andrea Morichetta, Publication Co-chairs, Dr. Suzan Bayhan, Dr. Sumitro Bhaumik, and Dr. Zhu Wang, Publicity Co-chairs, Dr. Symeon Papavassiliou, Panel Chair, and the website was handled by Md Sadman Siraj, and Panagiotis Charatsaris, Web Co-chairs. IEEE continuously supported the conference with Shirley P. Cisneros, ComSoc Project Manager, and Bruce Worthman, Financial Chair.
Introduction
The 24th IEEE International Conference on High Performance Switching and Routing (HPSR 2023) marked a significant pivot in networking research, focusing on Artificial Intelligent Next Generation (NextG) Integrated Communications and Computing Systems. As bandwidth becomes increasingly abundant, the industry is facing a new bottleneck: latency. The conference explored how ultra-reliable low latency (URLLC) is becoming the primary driver for emerging technologies like Extended Reality (XR), robotics, and haptic communications.
Modern network architectures face several critical challenges:
- The "In-Sequence" Constraint: Traditional switches are often over-engineered to ensure strict packet ordering, a requirement long thought essential for TCP.
- Satellite Hardware Limitations: Satellite networks require programmable parsers that are memory-efficient enough for space-constrained hardware.
- Dynamic Resource Management: Optimizing Service Function Chain (SFC) placement in fluctuating environments remains a complex spatial-temporal problem.
Technical Methodologies
Researchers presented several innovative approaches to address these NextG challenges:
- RACK-Aware Switching: Leveraging the Recent Acknowledgment (RACK) algorithm, Usubutun et al. demonstrated that modern TCP loss detection allows switches to reorder packets without performance degradation. This enables the use of simpler, more cost-effective load-balanced switch designs like the Birkhoff-von Neumann architecture.
- Multi-Table Programmable Parser (MTPP): This architecture for satellite networks utilizes five tables and protocol-independent fixed logic to represent parser graphs, drastically reducing storage requirements compared to standard PISA-based parsers.
- Double Deep Q-Network Decision Transformer (D3T): An AI-driven framework designed to optimize SFC placement by using transformers to predict and adapt to complex traffic patterns.
- Reconfigurable Multi-band Radios for CubeSats: A hybrid electronics-and-photonics design that enables small satellites to communicate across a massive spectrum range, from 2 GHz to 10 THz.
Key Insights and Ideas
A central theme of the conference was the transition from a bandwidth-centric to a latency-centric paradigm. Keynote speaker Dr. Shivendra Panwar argued that while 5G provides massive throughput, the true frontier for advances like wirelessly controlled robotics is reliable, sub-millisecond latency.
Furthermore, the conference challenged the long-standing "in-sequence" dogma of networking. By proving that modern TCP implementations are resilient to out-of-order arrivals, researchers have opened the door for high-speed, low-cost switch architectures that decouple data forwarding from strict ordering constraints.
Results and Recognition
The conference highlighted several breakthrough results:
- TCP Resilience: Experiments on wide-area core network testbeds showed that standard TCP implementations from the last decade perform nearly identically regardless of whether packets are reordered by load-balanced switches.
- Hardware Efficiency: The MTPP parser demonstrated significant storage savings, making advanced protocol parsing viable for resource-constrained CubeSats.
- Spectrum Versatility: New radio designs successfully addressed spectrum scarcity by covering millimeter-wave and Terahertz bands.
The Best Paper Award was presented to the team whose work on TCP reordering resilience promises to simplify future switch designs.
Future Directions
Despite these advancements, several open challenges remain. While low latency is achievable in controlled environments, maintaining it over wide-area cellular networks is still difficult. Additionally, the limited onboard processing power of CubeSats necessitates further research into offloading functionalities via SDN/NFV. Finally, the move toward "greener" data centers through programmable data planes is expected to be a major focus for future HPSR iterations.
Summary updated