Reproduce “Automating Multi-Tenancy Performance Evaluation on Edge Compute Nodes” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Joanna Georgiou, Moysis Symeonides, George Pallis, Marios D. Dikaiakos
An open-source automated benchmarking framework is presented for evaluating multi-tenancy performance across heterogeneous edge devices, revealing insights into workload co-location impacts on machine learning inference and energy consumption. The framework automates the deployment and comprehensive monitoring of diverse co-located workloads on real-world edge hardware.
Key insights
- Empirical evaluation is critical for edge deployments, as larger servers do not always guarantee superior performance or energy efficiency over smaller ones for specific workloads.
- Memory-intensive co-located workloads significantly increase machine learning service cold start times, highlighting the importance of careful workload placement in edge environments.
- The choice of machine learning execution backend and mode (local vs. streaming) drastically impacts performance and resource utilization, influencing the overall capacity for co-located services.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.