Predictive Analysis for Migration Schedulers
summary
Claims that using historical VM and process metrics to predict when processes complete lets the scheduler start live migration earlier, cutting pause time and using freed CPU/memory sooner. Proposed architecture uses a historian component + predictor that applies regression, SVM, clustering and anomaly detection and can use signal‑processing smoothing. Several prototype pieces exist but no full implementation; suggestions become early migration suggestions to the scheduler and require deployment and accuracy testing.
AI-written summary · about
chapters (4)
related talks
26mProactive Failure Prediction Through Anomaly Detection
14mPerformance Optimization in High-Volume Transaction Systems
25mSupercharging Pipeline Efficiency with ML Performance Prediction
56mAutomatically Scalable Computation
16mIntelligent Resource Scheduling for Co-located Latency-critical Services: A Multi-Model
22mEngineering Predictive Platforms
47mImprove Load Balancing With Machine Learning Techniques Based on the Sched_ext Framework
34mPatterns for building Highly Available Cluster Schedulers