Extensible Web Server Architecture


The Internet's ever-increasing popularity demands the Web sites to handle large amount of requests. This has created an urgent need for a more powerful Web server architecture to handle this problem. In this research, a distributed Web server architecture, called Extensible Web Server (EWS), is proposed. The EWS allows server nodes to be added or removed at any time. The dynamic nature of system configuration has added the challenges in the design of efficient document distribution algorithms to achieve high-performance Web services with minimum storage consumption and redirection overheads. Main features of EWS include : (1) Avoid storing unnecessary copy of document (2) Reduce the number of HTTP redirections. (3) Load Balancing.

Architecture of EWS

         The Extensible Web Server (EWS) consists of multiple distributed Web servers that function cohesively as a single logical web server. The main feature of EWS is that it allows server nodes to be added or removed at any time. In EWS, documents of a web site are partitioned and stored at different server nodes. Replication of popular document objects are permitted to improve the access locality and shorten the access latency.

        The EWS consists of two main subsystems: (1) Resource Monitoring Subsystem (RMS). (2) Web Service Subsystem (WSS). The RMS is the backbone of EWS, which is used to manage the computing and storage resources in EWS. The WSS consists of Web server software to manage the document distribution, document consistency, document migration, and handling client requests. Figure 1 shows an overview of the proposed EWS architecture.


Technical papers:

Ben. Ng and C.L. Wang, “Document Distribution Algorithm for Load Balancing on an Extensible Web Server Architecture,” First IEEE/ACM International Symposium on Cluster Computing and the Grid (CCGrid 2001), 15-18 May, 2001, Brisbane, Australia (Paper in Postscript, HTML)
Ling Zhuo, Cho-Li Wang, and Francis C. M. Lau, ``Load Balancing in Distributed Web Server Systems with Partial Document Replication,'' The 2002 International Conference on Parallel Processing (ICPP-2002), Vancouver, British Columbia, Canada, August 18-21, 2002. (Draft in PDF)

Research Students: 

Ben NG (Graduated)
Ling Zhuo (now with USC)
Chen Weisong (2002, M.Phil)