Detecting and Understanding Performance Failures with Course-grain Monitoring

Performance failures are commonplace in most computing environments; without system monitoring they typically go undetected. The performance monitoring project uses low level system monitoring combined with an infrastructure to centralize the information tracks long term system behavior. With the data visualization tools we can detect performance failures, track resource utilization trends, and document intermittent failures. The monitoring infrastructure allows comparisons between multiple applications in a real production environment, verification that upgrades/changes improved performance, performance experiments and benchmarking applications. The infrastructure is used as a too to evaluate the characteristics of specific systems/applications and for research on automatic failure analysis and diagnosis.