MétaCan
Menu
Back to cohort
Record W3161148246 · doi:10.1145/3440757

Predicting Performance Anomalies in Software Systems at Run-time

2021· article· en· W3161148246 on OpenAlex

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.

Bibliographic record

VenueACM Transactions on Software Engineering and Methodology · 2021
Typearticle
Languageen
FieldComputer Science
TopicSoftware System Performance and Reliability
Canadian institutionsIBM (Canada)Thompson Rivers UniversityQueen's University
Fundersnot available
KeywordsComputer scienceBaseline (sea)Precision and recallAnomaly detectionSoftwareRecallAnomaly (physics)Data miningSoftware systemMachine learningReal-time computingArtificial intelligenceOperating system

Abstract

fetched live from OpenAlex

High performance is a critical factor to achieve and maintain the success of a software system. Performance anomalies represent the performance degradation issues (e.g., slowing down in system response times) of software systems at run-time. Performance anomalies can cause a dramatically negative impact on users’ satisfaction. Prior studies propose different approaches to detect anomalies by analyzing execution logs and resource utilization metrics after the anomalies have happened. However, the prior detection approaches cannot predict the anomalies ahead of time; such limitation causes an inevitable delay in taking corrective actions to prevent performance anomalies from happening. We propose an approach that can predict performance anomalies in software systems and raise anomaly warnings in advance. Our approach uses a Long-Short Term Memory neural network to capture the normal behaviors of a software system. Then, our approach predicts performance anomalies by identifying the early deviations from the captured normal system behaviors. We conduct extensive experiments to evaluate our approach using two real-world software systems (i.e., Elasticsearch and Hadoop). We compare the performance of our approach with two baselines. The first baseline is one state-to-the-art baseline called Unsupervised Behavior Learning. The second baseline predicts performance anomalies by checking if the resource utilization exceeds pre-defined thresholds. Our results show that our approach can predict various performance anomalies with high precision (i.e., 97–100%) and recall (i.e., 80–100%), while the baselines achieve 25–97% precision and 93–100% recall. For a range of performance anomalies, our approach can achieve sufficient lead times that vary from 20 to 1,403 s (i.e., 23.4 min). We also demonstrate the ability of our approach to predict the performance anomalies that are caused by real-world performance bugs. For predicting performance anomalies that are caused by real-world performance bugs, our approach achieves 95–100% precision and 87–100% recall, while the baselines achieve 49–83% precision and 100% recall. The obtained results show that our approach outperforms the existing anomaly prediction approaches and is able to predict performance anomalies in real-world systems.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.001
metaresearch head score (Gemma)0.001
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesMeta-epidemiology (narrow)
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: Simulation or modeling
GenreCandidate signal: Methods · Consensus signal: Methods
Teacher disagreement score0.214
Threshold uncertainty score1.000

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0010.001
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.001
Science and technology studies0.0000.000
Scholarly communication0.0000.000
Open science0.0000.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.038
GPT teacher head0.262
Teacher spread0.224 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it