gridMonSteer: Generic Architecture for Monitoring and Steering

Grid environments present a virtual black box to scientists running legacy applications, with the Grid infrastructure effectively hiding the running application on a resource over which the scientist generally has limited or no control. Existing moni-toring systems that allow Grid-enabled applications to communicate their progress and receive steering information are inapplicable as they require code modification, and this not possible in true legacy scenarios. While a black box approach may be ac-ceptable in batch execution scenarios, it means the Grid is cut off to legacy applications where inter-actions or intermediate results are required. In this paper we present gridMonSteer, a sim-ple, non-intrusive architecture that allows scien-tists to receive intermediate results and interact with legacy applications running in a Grid envi-ronment. This architecture is Grid middleware in-dependent and can be employed to monitor ap-plications submitted via any Grid resource man-ager (e.g. GRAM, GRMS or Condor/G) or run-ning within a Grid service framework. We present a case study describing how gridMonSteer enables legacy applications to act as active components in Grid workflows, dynamically driving and steering the workflow execution.