The proliferation of latency and safety-critical applications such as AR/VR, surveillance, autonomous vehicles, and health monitoring has forced service providers to move from the distant cloud closer to users; namely, to the edge. Edge computing strives to provide a better quality of experience by seeking primarily faster response times. Our team has been surfing on this new trend by exploring cutting-edge programmable network devices (e.g., programmable switches, SmartNICs) as processing units for edge-based applications. The rationale is simple: these devices are closer to the user than any server. Ultimately, our goal is to enable the next generation of low-latency and high-throughput applications to meet the stricter than ever user requirements.
Towards Network-accelerated ML-based Distributed Computer Vision Systems Hisham Siddique, Miguel Neves, Carson Kuzniar, Israat Haque IEEE International Conference on Parallel and Distributed Systems (ICPADS) 2021
NetPixel code is available on github here
israat at dal dot ca
mg478789 at dal dot ca
Carson.Kuzniar at dal dot ca
hisham.siddique at dal dot ca