Spring 2002

Due Date: January 22, 2002 by midnight.

The primary goal of this project is to build a multithreaded application that requires sufficient programming effort to get you ready for future projects that will be assigned in the class. You will also learn how to use the Pthreads library to create a multithreaded application. This library is available on most platforms.

The specific goal of the project is to build a highly simplified web server. Clearly, a web server handles requests that come from remote clients. In this project, you do not need to worry about remote nodes and distributed programming. However, since a web server handles requests for documents stored at it using the hyper text transfer protocol (http), you will have look up http and the messages that can be exchanged using this protocol between a client and the web server. Your server will read such requests from a file. Each request will have all the information associated with it. Also, documents on the server are typically stored on disk. You will make use of a large file to simulate the disk. Hence, disk operations will be translated to read and write requests to the ?disk file?.

The server will be multithreaded and it will perform a number of functions to efficiently provide the documents that are requested by clients. The server will consist of the following functions, each of which will be implemented by one or more threads.

A dispatcher thread will read client requests. A delay should be introduced between reading client requests to mimic the asynchronous arrival of client requests.
The dispatcher thread will hand off requests to a resolver thread. This thread will translate URLs to local document identifiers. The local identifiers can be assumed to be between 0 and n.
Once requests are translated to local document identifiers and other checks are done, the resolver will add the requests to a work queue that contains pending requests. A number of worker threads will process these requests.
The server will employ a cache of recently accessed documents to improve performance. A cacher thread will implement an approximate LRU replacement algorithm to decide what documents are kept in the cache. You can set the cache size so that the fault rate is low (e.g., below 20%).
You can employ one or more GUI threads that show the contents or other information and requested documents are returned. These threads will take actions similar to what a client will do with the received document.

You can also use additional threads if that helps you structure the server better or to provide better performance or scalability.

Obviously, to successfully complete the project, you must understand how threads work and are synchronized using locks and condition variables. When it is better to block a thread rather than busy wait, you should make use of condition variables.

To test the program, the TA will use an input file that is a sequence of http requests. Other information such as URL names etc. will be provided to you next week. You should generate suitable output to demonstrate that your implementation handles the requests according the specification of the project.

New added

The input file is provided by yourself when you turn in your project. For example, the format will look like

GET /path/yourfile1 HTTP/1.0

PUT /path/yourfile2 HTTP/1.0 sourcefileToPut

You don't need to zip the sourceToPut files in your turn-in. But in the input file you turn in, please give the sourcefileToPut file the required name like 1.txt, 2.txt, 3.txt, and so on. So I will provide the content of those sourcefileToPut.

http1.0 specification for your reference

In a report that needs to be submitted with an online copy of all code and test files, you need to briefly discuss the threads that are crated by your implementation, the major data structures and how they are shared, and details of the code that is executed by each thread. The TA will provide details of how the projects should be submitted.

TURN-IN:

This is one minute prior to midnight on Tuesday night. No late assignments will be excepted

unless prior arrangements have been made.

Assignments will be submitted by e-mailing a tar archive of your work to the account

cs6210@cc.gatech.edu. Instructions for creating a tar file will be posted to the newsgroup. Make

sure that you do NOT include any binaries in your tar files. I am going to compile your programs

from scratch. If you have any special commands needed to compile your code then mention them

in your README file.