PERFORMANCE AND PARALLEL PROCESSING TEAM

PP1

RESEARCH TASK: Testbed Performance
PLANNED ACTIVITY
Address parallel I/O issues in dealing with browsing of wavelet-based large images. It requires us to re-examine the decomposition of large images and study the disk model and its impact on the design of data placement.
ACTUAL ACTIVITY
We have addressed this issue by using clustering support for an ADL server. We find that clustering workstation nodes together is an economic approach to improving the performance. We have also studied the parallel I/O issues for wavelet image placement and did not find substantial performance improvement. Thus we decided to focus on clustering and load balancing support.

PP2

RESEARCH TASK: scalable Web server/client technology
PLANNED ACTIVITY
Address issues of scalable web client/server technology. In particular, develop a Java-based implementation of the SWEB++ system for adaptive client-server computing; analysis of SWEB++ performance model and scheduling algorithms; and development and verification of runtime load monitoring and bandwidth measurement.
ACTUAL ACTIVITY
We have addressed this issue by developing the SWEB++ system for adaptive client-server computing and a scheduling and partitioning scheme that considers both resource availability at a multi-processor server and a client. We have also developed a Web caching scheme (the Swala project) to improve the Web server cluster performance by cooperative caching.

PP3

RESEARCH TASK: Applications of Adaptive Techniques
PLANNED ACTIVITY
To continue to work applications of adaptive techniques. One is to improve the current wavelet-based image browsing algorithm design and implementation. Another is to examine the client-server interaction in other document browsing and searching situations.
ACTUAL ACTIVITY
We have applied the client-server techniques for document and image browsing. In particular, we have developed a Java package for progressive wavelet image browsing.