Daphne: performance debugging using model-driven anomaly characterization

Authors:
Christopher Stewart;Kai Shen
Affiliations:
University of Rochester;University of Rochester
Venue:
Proceedings of the twentieth ACM symposium on Operating systems principles
Year:
2005

Citing 3
Cited 0

Performance debugging for distributed systems of black boxes

SOSP '03 Proceedings of the nineteenth ACM symposium on Operating systems principles
Path-based faliure and evolution management

NSDI'04 Proceedings of the 1st conference on Symposium on Networked Systems Design and Implementation - Volume 1
Performance modeling and system management for multi-component online services

NSDI'05 Proceedings of the 2nd conference on Symposium on Networked Systems Design & Implementation - Volume 2

Quantified Score

Hi-index	0.00

Visualization

Abstract

It is not uncommon for complex systems (like multi-component online services) to perform worse than expected. Configuration errors, buggy code, and tricky middleware/OS behavior induce performance anomalies under certain workloads and system configurations. It is challenging to pinpoint root causes of such performance problems in complex systems. Currently, performance problems are solved by system administrators who intimately understand the system and are willing to forgo a personal life.