- Scalable Parallel File System and Qos-driven Storage Management for High-end Computing Systems
Parallel storage systems are the backbone of high-end computing (HEC) systems. With the ever-increasing demand from HEC applications and the ever-increasing size of HEC systems, current parallel storage systems are becoming insufficient to support their Quality of Service (QoS) and scalability needs. This project studies novel parallel file system and storage management techniques to address these challenges.
- Autonomic Resource Management of Virtualized Computing Systems
On-demand computing with networked data-centers poses challenging problems of delivering expected QoS and doing so with acceptable management complexity. This project studies autonomic architectures and mechanisms for cooperative resource allocation and QoS delivery by virtualized data centers, in which application services are provided in virtual-machine based execution environments. [More]
- Grid Virtual File System
GVFS is a virtualized distributed file system that provides high-performance and secure data access in grid/wide-area environments, and allows for seamless integration with unmodified applications. GVFS helps users access grid data in the same way as using traditional LAN file systems. It is also an integral part of the In-VIGO grid virtualization middleware. [More]
- Application-Tailored Grid Data Management with WSRF-based Services
This project focuses on novel data management middleware that leverages Web Service Resource Framework to schedule, customize and monitor dynamic Grid data sessions. Application-tailored session configurations enable the selection of both performance-related features (partial/whole-file transfers, cache and consistency models) and reliability features (copy-on-write checkpointing, replication, failure detection and data access redirection). [More]
- Data Provisioning for Grid Computing on Virtualized End-Resources
Grid computing with virtual machines promises the capability of provisioning a secure and highly flexible computing environment for users. This project develops middleware to provide efficient data management service for grid VMs (for both VM states and user data). It can be applied to VMs of different kinds (VMware, Xen, UML etc.). [More]
- Dynamic Policy Disk Caching for Storage Networking
Proposed client-side disk caching to solve the scalability problem for storage networking. Designed and implemented the dm-cache kernel module, a generic block-level disk cache that supports transparent caching for different types of storage networks and enables policy-guided dynamic cache customizations. [More]
- Cyberinfrastructure for Dynamic Data-driven Brain-machine Interfaces (DDD-BMI)
The goal of DDD-BMI is to model brain from experiments with live subjects and to design brain-inspired assistive systems. This project develops a middleware-based cyberinfrastructure to support this time-critical and resource-demanding application. The resulting cyber-workstation enables BMI experiments to be conducted in a closed-loop manner, including in vivo brain signals acquisition, reliable network messging, parallel BMI computing, and real-time robot control. [More]
- Fine-Grain Grid Data Management and Applications in LSS Medical Imaging
This project aims to improve the state-of-the-art in data management techniques for grids to allow seamless and high-performance integration of data generated by medical devices with distributed computers. It enables the deployments of network-computing based medical applications (e.g. Light-Scattering Spectroscopy analysis) for early cancer detection.
- Parallel Computing on Heterogeneous Distributed Clusters
Coupling clusters in grid environments is important to harness existing resources for large scale problem solving and multi-institutional cooperation. In this project a MPI-coupling technique (PACX-MPI) and a grid-wide file system (GVFS) are combined to support parallel computing on distributed, heterogeneous and cross-domain clusters.