A WWW-based in-class response system

Abstract

Using the current network infrastructure and the Internet capabilities, a WWW-based in-class-response system is developed from which the instructor can obtain instant feedback about the student's comprehension level during lecture. Using a Web browser in class as a form of instructor-student interaction, students can improve communication, shorten response time to questions and increase overall learning. Immediate access to the students' response to questions posted in-class will give the instructor a chance to initiate early corrective action and enhance overall teaching performance. One of the concerns resulting from the use of such a system in an in-class setting, however, arises due to the potential problem of server overload. With a classroom size of 50 to 100 students working within the same one to two minute window, there may be Web server performance problems caused by the large number of submissions within such a short time. Observing the unsatisfactory performance of the Web Server during the initial stages of this research, led to the implementation to the proposed solution to overcome this problem. Implementation of the FastCGI interface on the Web Server is used to minimize response time. A comparison between CGI and FastCGI shows the performance improvement which stems from having a configurable number of processes running on the Web Server to serve the concurrent requests. A uniformly distributed scenario for an in-class response setting (e.g., having 100 students accessing the server in a 30 seconds window), would require a service rate from the server of browser's results, having a response time of between 12 and 17 seconds per connection using a fraction of the expected load, leads to the conclusion that the original testbed based on a SPARC I will not be able to sustain the expected load for an in-class setting. A much lower response time and a higher throughput was obtained after installing the CGI programs on a Ultra Sparc 2. The benchmark tests using 84 Webclients showed a throughput of 17 and 194 connections per second using the CGI and FastCGI access method, respectively. The Web browser tests showed an average response time of less than one second per connection using a higher load than the expected 100 requests in a 30 second window. These results leads us to conclude that the Web Server based on the Ultra Sparc 2 will be able to handle the expected in-class load, at the much higher expense of the latter machine.