Operational performance metrics in a distributed system. Part I.
Bert N. Corwin, Robert L. Braddock · 1992
This paper presents a strategy for using operational performance metrics to characterize and monitor distributed systems.Spe@ metrics are discussed and three target audiences are identified: Operations Managers, Operatwrrs Personnel, and Administrators.By using this strategy, these metrics assist in managing large, complex distributed environments as integrated systems.The performance metrics captured during system modeling and development are contmrted to these operational pe~ormance metrics.systems,from specific nodes, such as a file .swer, to large, integrated systems.Nearly all of these studies propose the definition of system workload models, followed by rdloeation of these models to proposed distributed systems to provide projections of system throughput and availability.The focus of these studies has been on the development-side of systems.Once a distributed system is implemented, a strategy is needed to monitor the performance of the system and to provide for future capacity planning.Operational performance metrics for centralized systems have been extensively studied [5, 6]; however, little work has been reported on monitoring the performance of operational distributed systems in an integrated manner.This paper proposes operational performance metrics for distributed systems based on thrw key factors: a definition of the target audience for the metrics, appropriate metrics