The University of Arizona, Department of Computer Science

Web Server

Overview

We are building an appliance that stores and serves web pages. It is accessed via HTTP and stores web pages in memory and on local disk. It can function in three ways. First it can be configured as a web cache, in which case it retrieves missing pages using HTTP to connect to a remote web server. Secondly it can be configured as a web server without local storage, in which case it retrieves missing pages by using NFS to access a local file server. In the third case it can be configured as a web server with local storage, in which case it retrieves pages from local disk. The advantage of using the Scout web server in the last two scenario is that web pages can be laid out on the local disk in a manner that is optimized for HTTP access patterns, rather than according to a traditional (Unix-like) file system. In all cases many denial of service attacks can be prevented by using Escort, the comprehensive security architecture for Scout

Status

• The current implementation supports HTTP 1.0. It uses NSF or the local disk to access web pages from the local file system.
• The web server has been implemented in the current Scout version and using Escort -- the comprehensive security architecture for Scout -- to prevent denial of service attacks.
• All Scout-related pages can be accessed via a Scout web server by cliking here

Papers

[Spat98b] O. Spatscheck, L. Peterson. Defending Against Denial of Service Attacks in Scout.Proceedings of the Third Symposium on Operating System Design and Implementation. February 1999.

Contributors

• Oliver Spatscheck, Graduate Student
• Nick Ogurtsov, Graduate Student