A Kernel-Based Approach for Accurate Steady-State Detection in Performance Time Series

2025-06-04

AI Summary

Accurately identifying the transition point from warm-up to steady state in performance metric time series is critical for improving benchmark accuracy and reproducibility. This paper introduces, for the first time, steady-state detection principles from chemical process engineering into systems performance analysis. We propose an online step-change point detection method that integrates kernel-based smoothing, sliding-window statistical hypothesis testing, and adaptive thresholding—enhancing robustness against noise and irregular temporal patterns. Compared with the state-of-the-art approaches, our method reduces total error by 14.5% and significantly improves the precision of steady-state onset localization. It effectively mitigates performance evaluation bias caused by premature or delayed steady-state declarations. The framework enables automated, high-fidelity benchmarking, establishing a new paradigm for reliable system performance assessment.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Time-Series/Data Streams
• Intelligent Robots: State Estimation

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

This paper addresses the challenge of accurately detecting the transition from the warmup phase to the steady state in performance metric time series, which is a critical step for effective benchmarking. The goal is to introduce a method that avoids premature or delayed detection, which can lead to inaccurate or inefficient performance analysis. The proposed approach adapts techniques from the chemical reactors domain, detecting steady states online through the combination of kernel-based step detection and statistical methods. By using a window-based approach, it provides detailed information and improves the accuracy of identifying phase transitions, even in noisy or irregular time series. Results show that the new approach reduces total error by 14.5% compared to the state-of-the-art method. It offers more reliable detection of the steady-state onset, delivering greater precision for benchmarking tasks. For users, the new approach enhances the accuracy and stability of performance benchmarking, efficiently handling diverse time series data. Its robustness and adaptability make it a valuable tool for real-world performance evaluation, ensuring consistent and reproducible results.

Problem

• Accurately detect transition from warmup to steady state in performance time series
• Avoid premature or delayed steady-state detection in performance analysis
• Improve accuracy of phase transition identification in noisy time series

Innovation

• Kernel-based step detection for steady-state identification
• Window-based approach enhances phase transition accuracy
• Combines statistical methods for robust online detection