Predicting Timing Failures in Web Services

Abstract

Web services are increasingly being used in business critical environments, enabling uniform access to services provided by distinct parties. In these environments, an operation that does not execute on due time may be completely useless, which may result in service abandonment, and reputation or monetary losses. However, existing web services environments do not provide mechanisms to detect or predict timing violations. This paper proposes a web services programming model that transparently allows temporal failure detection and uses historical data for temporal failure prediction. This enables providers to easily deploy time-aware web services and consumers to express their timeliness requirements. Timing failures detection and prediction can be used by client applications to select alternative services in runtime and by application servers to optimize the resources allocated to each service.