Complex high-capacity training and disaggregated inference workloads are being scaled-across multiple sites, blurring the lines between contemporary scale-out and scale-up interconnection stacks. All the classical issues, such as jitter, bursts, flow/congestion, long-tail latencies, etc., are amplified over long distances, requiring deliberate interventions across the full interconnection stack: algorithms, operational tools, communication frameworks, and network fabrics. This edition of Hot Interconnects will explore these interventions behind operationalizing interconnects over long, short, and shorter distances.
Call For Papers
HotI 2026 solicits original research papers addressing the design, analysis, implementation, and evaluation of high-performance interconnects and closely related technologies. Topics of interest include, but are not limited to, interconnect architectures, network protocols, networking software, communication runtimes, congestion control, optical and emerging fabrics, accelerator and chiplet interconnects, disaggregated systems, and system–network co-design.
Paper submissions are expected to open in March 2026. Like the previous years, we will consider regular and hot-topic papers. The full Call for Papers, including important dates and submission guidelines, will be released closer to the submission opening.
Submissions
| Event | Date |
|---|---|
| Paper abstract deadline | TBD |
| Submission deadline | TBD |
| Notification of acceptance | TBD |
| Camera-ready | TBD |
Topics of Interest
- Systems software for communication
- Novel and innovative interconnect architectures
- Accelerator interconnects, e.g. NVLINK, Infinity Fabric
- Network software/hardware designed for AI/ML workloads
- Multi-core processor interconnects
- System-on-Chip Interconnects
- Chiplet-interconnect technologies such as UCIe and BOW
- Advanced chip-to-chip communication technologies
- Optical interconnects
- Protocol and interfaces for inter-processor communication
- Survivability and fault-tolerance of interconnects
- High-speed packet processing engines and network processors
- System and storage area network architectures and protocols
- High-performance host-network interface architectures
- High-bandwidth and low-latency I/O
- Pb/s switching and routing technologies
- Innovative architectures for supporting collective communication
- Novel communication architectures to support cloud & grid computing
- Centralized and distributed cloud interconnects
- Requirements driving high-performance interconnects
- Traffic characterization for HPC systems and commercial data centers
- Software-defined networking and software overlay networks
- Software for network bring-up, configuration and performance management (OpenFlow, OpenSM)
- Data Center Networking