English

Profiling Multi-Level Operator Costs for Bottleneck Diagnosis in High-Speed Data Planes

Signal Processing 2025-08-14 v1

Abstract

This paper proposes a saturation throughput delta-based methodology to precisely measure operator costs in high-speed data planes without intrusive instrumentation. The approach captures non-linear scaling, revealing that compute-intensive operators like CRC exhibit super-linear behavior, while most others are sub-linear. We introduce the Operator Performance Quadrant (OPQ) framework to classify operators by base and scaling costs, exposing a cross-architecture Quadrant Shift between Arm and x86. This method provides accurate, architecture-aware bottleneck diagnosis and a realistic basis for performance modeling and optimization.

Keywords

Cite

@article{arxiv.2508.09574,
  title  = {Profiling Multi-Level Operator Costs for Bottleneck Diagnosis in High-Speed Data Planes},
  author = {Zhiyuan Ren and Yutao Liu and Wenchi Cheng and Kun Yang},
  journal= {arXiv preprint arXiv:2508.09574},
  year   = {2025}
}
R2 v1 2026-07-01T04:47:40.802Z