Monitoring Virtualized Infrastructure in the Context of Grid Job Execution

Abstract—This paper describes a new direction in the develop-ment of the Logging and Bookkeeping service, a gLite component tracking job life cycles in high performance computing grids. From its early days, Logging and Bookkeeping is used to track not only jobs themselves, but also the wider details of the job execution environment. Since a large portion of the grid infrastructure is now virtualized, the work at hand concerns tracking the virtualized nature of that runtime environment. With virtualization and cloud technologies being highly flexible and dynamic, the authors believe that it is very important to gather and keep status information for virtual machines used to run the workload. A newly defined monitoring entity – a virtual machine – is integrated with job state information and provides an enhanced view of the current state and history of both the computing job and the underlying infrastructure, as well as their mutual relationship. This paper explains the motivation and discusses the architecture of the newly emerged solution for monitoring virtualized resources – uniquely – in the same context as the workload they are processing. Keywords-grid; cloud; virtualization; monitoring; relationship I