Research Progress and Results • In the last several months of the project (during the no-cost extension), we have investigated providing parallel data processing services on distributed multi-core workstations. In particular, we explored an innovative hybrid architecture for data-intensive services on top of unreliable environments, by supplementing volatile nodes with a small group of dedicated, highly available nodes. We studied the impact of such an architecture on the data replication

more »

... ement, as well as the effectiveness of hybrid-resource-aware data processing task scheduling. Our results, obtained from Virginia Tech cluster using both synthetic node availability traces and a real-world desktop grid trace show promising results. We are submitting a technical paper on this study for conference publication.