My research interests lie in the area of parallel and distributed computing with the emphasis on software supports and optimization techniques for data-intensive applications on parallel and distributed systems. We have addressed replica and server placement problems in various cluster and Grid environments to improve reliability and performance of the systems. We have studied replication transition problems so that the replica placement can adapt to new user preference and system configuration. With multiple servers in the systems, we have proposed efficient algorithms to improve the aggregate bandwidth and reliability of data transfers. Based on the above results, we have designed and implemented a high-performance virtual storage system for Taiwan UniGrid. We have also investigated resource management problems and proposed bidding based mechanisms for resource selection.
Recently, we focused our attention on cloud computing. Cloud computing is a new and promising paradigm in which dynamically scalable and often virtualized resources are provided as a service over the Internet. It provides many advantages, such as resource sharing and scaling, reduced capital expenditure, simplified IT management, higher resource utilization, and better reliability. However, there are still many challenging issues needs to be explored. Now we are investigating distributed file systems for cloud computing and cloud middleware for highly interactive application services. The success of GFS motivates us to explore adaptive data and storage management for different workloads and technological environments. This demands innovative solutions to challenging issues, such as workload profiling and clustering, inter-file association, environment benchmarking, data and storage management strategies for different workloads and environments, on-line monitoring and workload classification, and management strategy selection. The popularity of networked virtual environments demands resource-efficient cloud middleware for highly-interactive application services, with the goal to support massive multiplayer NVEs, improve efficiency and power usage on servers, and guarantee real-time interactivity. This demands innovative solutions to challenging issues, such as performance assurance and fairness in virtualization, power/resource management, workload prediction, fast live migration, failure handling, in-memory database design, and dynamic data partitioning/replication/localization.