ProvenanceLens: Service provenance management in the cloud

Service provenance can be defined as a profile of service execution history. Queries of service provenance data can answer questions such as when and by whom a server is invoked? which services operate on this data? What might be the root cause for the service failure? Most of the organizations today collect and manage their own service provenance in order to trace service execution failures, locate service bottlenecks, guide resource allocation, detect and prevent abnormal behaviors. As services become ubiquitous, there is an increasing demand for proving service provenance management as a service. This paper describes ProvenanceLens, a two-tier service provenance management framework. The top tier is the service provenance capturing and storage subsystem and the next tier provides analysis and inference capabilities of service provenance data, which are value-added functionality for service health diagnosis and remedy. Both tiers are built based on the service provenance data model, an essential and core component of ProvenanceLens, which categorizes all service provenance data into three broad categories: basic provenance, composite provenance and application provenance. In addition, ProvenanceLens provides a suite of basic provenance operations, such as select, trace, aggregate. The basic provenance data is collected through a light-weight service provenance capturing subsystem that monitors service execution workflows, collects service profiling data, encapsulates service invocation dependencies. The composite and application provenance data are aggregated through a selection of provenance operations. We demonstrate the effectiveness of ProvenanceLens using a real world educational service currently in operation for a dozen universities in China.