Our paper "Pruning Edge Research with Latency Shears" has been accepted at ACM HotNets 2020! This work provides a reality check for edge computing research by measuring actual latencies achievable in real-world edge deployments.
The edge computing hype often promises millisecond-level latencies, but our measurements reveal that network latency remains the dominant factor and edge proximity helps less than commonly assumed. Many papers in this space make overly optimistic assumptions, ignoring network propagation delays, routing realities, and last-mile access network latencies. Our research challenges these unrealistic expectations and encourages the community to use honest baselines and measurement-driven evaluation.
HotNets is the premier venue for presenting provocative early-stage ideas in networking, encouraging bold positions that challenge conventional wisdom. The paper sparked important discussions about when edge computing truly makes sense and the importance of realistic modeling in research.
This research has influenced our subsequent work on Oakestra and continues to guide our approach to building practical edge systems with realistic performance expectations. Several years later, the field has matured and realistic assessments of edge computing are more common—partly due to work like this.
Conference: ACM HotNets 2020
Video: YouTube
Related: