Every day the Web gains more importance for consumers and corporations alike. However, while access to Web-based networks brings tremendous value, the number one complaint among Web users is how long it takes to download Web pages. Whether those pages are delivered from informational Web sites or enterprise applications, the delays are unacceptable.

Throwing bandwidth at the problem will not help significantly. Slow Web pages are caused by a combination of distance between users and servers, and congestion of heavily trafficked networks beyond the control of any one organization.

The only possible solution for fast Web page response time is to move the data closer to end users. Network caching, which stores fresh copies of the most frequently requested Web objects near groups of users, can deliver Web pages with the same response time people have come to expect from traditional applications.

Within the scope of my degree work, I designed and implemented a caching proxy server that handles HTTP traffic. The proxy features:

• A robust caching mechanism, that is optimized to minimize the time that end users wait for documents to load by giving preference to documents that must be loaded over the slowest Internet links, that are from servers that take long time to connect to, that have been referenced more frequently, and that are small. This is achieved by incorporating estimations of bandwidths between the local server and remote servers, estimations of connection times to remote servers, frequency of references to cached documents, and cached document sizes into the cache removal algorithm. The weights of the parameters are all fully customizable.
• Multithreading, to allow concurrent handling of several HTTP requests. Great care has been given to concurrency control, to ensure mutual exclusion between threads where needed and to avoid deadlocks.
• Strong initialization and shut-down routines. These enable recovery of all data (cached documents, servers statistics, configuration of the proxy server) upon restart that follows a proper shut-down, and of all consistent data upon restart that follows a crash.
• The ability to define a list of servers whose documents will not be cached by the proxy, and the ability to define a proxy to which all unfulfilled requests will be forwarded.
• A graphical user interface, that features a log area, a cache hits-misses pie chart and a cache occupancy chart. Customization of the proxy is done via the graphical interface.