Benchmarking in The Dark: On the Absence of Comprehensive Edge Datasets
summary
Paper demonstrates that available datasets miss key edge attributes and builds a concrete mitigation: compose attributes from multiple public sources to synthesize realistic edge storage workloads. Method merges NYC taxi pickup/dropoff mapped to zones, hotspot locations mapped to zones and Wikipedia click sessions to model object requests to produce IO traces with time, location, node and object fields. Results show time-varying, nonuniform demand and partial correlations that justify the composition for evaluating edge storage designs.
AI-written summary · about
chapters (8)
- 0:00Intro
- 0:23MOTIVATION
- 2:17ESTING WORKLOADS
- 3:28DATASETS AND ATTRIBUTES
- 4:56WORKLOAD COMPOSITION
- 6:57CHARACTERCING THE SYSTEM AND ITS USERS
- 8:27GENERALIZATION
- 9:36SUMMARY
related talks
40mSynthetic data: when, why, and how
14mVector Search Benchmark[eting]
07mData Gravity and Kubernetes: Managing Large-Scale Data Ingest With Minimal Latency
50mBig Data on Smallish Machines
32mLarge-Scale Graph Processing on Emerging Storage Devices
27mData @scale Seattle - Yucheng Low
22mWrite-Through Object Cache for Fleets of Global Edge Clusters
31mFinding the Needle in the Haystack: Predicting Storage Device Failures in...