Sponsored by U.S. National Science Foundation under grants CDA-9619900, EIA-9729828, CCR-9988266, ACI-0203592, CCF-0611750, DMS-0624849, CRI-0708232, CNS-0702488, NASA under grant 03-OBPR-01-0049, and Sun Microsystems AEG Grant(07).

Current Projects

MAPS: system support for mobility and adaptation in pervasive services

With wide spread of wireless access networks and permeation of mobile devices, the Internet continues to push its envelop and merge its front tier into the physical world we're living. A key requirement for such pervasive servicdes is accessiability anytime, anywhere on any devices with quality assurance. A challenge is that clients may access the services on a great diversity of devices (from cellular phone, PDA, to PC) via different types of access networks (from wireless to broadband). For service quality assurance, the services must be differentiable according to the client capacity and network bandwidth. In this project, we focus on resource constrained service differentiation techniques for media rich Web contents and streaming applications.

    References:
  • Harmonic resource allocation for differentiated QoS on streaming servers, IEEE Trans. on PDS, Sept. 2004
  • Sematics-based prefetching in Internet services, IEEE Trans. on DKE, May 2004.
  • Personalized email management at network edges, IEEE Internet Computing, March/April 2005.
  • NaproxI: a hash based collaborative proxy system for content adaptation, Proc. of ICDCS Workshops, 2005

Mobile agents for data-intensive network applications and security

Agent is a special type of software component that has autonomy. It behaves like a human agent, working for clients in pursuit of its own agenda. A mobile agent has the ability to travel from machine to machine autonomously, carrying its code, data, and running state. Mobile agents grew out of earlier technologies like mobile objects and process migration. Early mobility research was mostly targeted at ™closedš distributed systems, in which programmers have complete knowledge about the system and full control over the disposition of the machines. By contrast, agents are tailored to open and distributed environments because of its autonomy property. This project focuses on developing a secure and flexible mobile agent middleware system and a novel agent-oriented programming paradigm in support of increasingly important network-centric applications.

    References:
  • Naplet: A secure and flexible mobile agent framework for network-centric applications
  • An integrated framework for network management using mobile agents

Application-Level Scheduling for High Performance Computing on Clusters

With the advance of fast communication network and the popularity of multiprocessor servers, clusters of servers are emerging as one of the most cost-effective ways for high performance computing. Delivering their full potential heavily relies on efficient cluster-wide resource management systems. This on-going project focuses on run-time scheduling techniques for irregular problems on SMP clusters. Since the cluster deploys a hierarchy organization, it is the depth of the memory hierarchy with its different access primitives and costs at each level that makes scheduling challenge. SMP clusters tend to be run in a multiprogramming model. The scheduling techniques must also be robust against interference of parallel jobs.

    References:
  • Stochastic remapping of dynamic bulk synchronous computations in distributed systems
  • Distributed Shared Arrays: An integration of message passing and multithreading on SMP