The need for service providers to offer SLAs to gain consumers confidence has become increasingly apparent. Our work focuses on providers who manage the execution of long-running data-driven grid workflows which are often characterized by user specified deadline constraints. Service providers responsible for managing the execution of such workflows tend to overprovision system resources to reduce the possibility of deadline violations. Further, uncertainty in reliably estimating the QoS parameters makes advanced reservation mechanisms less feasible. The paper proposes an alternate real time management strategy that supports just-in time resource allocation for executing workflows. Towards this end, the paper outlines a scalable and responsive architecture that incorporates policy based management components for allocating resources and integrates monitoring into the workflow itself. The proposed architecture allocates additional resources from the grid pool to the data-parallel operati...