Problem
Suppose you are interested in comparing network performance for various network bandwidths consisting of 100 Kbps, 500 Kbps, 5 Mbps, and 5 Gbps. In this experiment, you decide to visit the course webpage for one of your classes. The course page you selected consists of the base page (which is 10 Kbytes long), along with 5 embedded objects, each of which also consists of 10 Kbytes. To download this information, assume that you use a TCP variant with an MSS of 1460 bytes (a common size in many implementations).
1. Assume a round-trip time of 50 ms. Construct a chart that compares the total response time for retrieving the web page and its embedded objects under a pipelined-persistent, and under a non-persistent HTTP connection.
2. Repeat (1) but assume an RTT of 1 second.
3. Now consider using the most common variant of HTTP: parallel non-persistent connections. Let p denote the maximum number of concurrent parallel connections the browser is permitted to open to a given server. Before opening any parallel connections, the browser must first open an HTTP connection to retrieve the base page. Upon receiving and parsing this file, the browser will then initiate p parallel connections to retrieve the embedded objects. Given this behavior, show that the total response time for the download of the web page and its contents is equivalent to:
2(E/p + 1)RTT + (E+1)O/C + SSL
where E is the number of embedded objects, O is the size of an embedded object (which for this problem can be assumed as constant), C is the capacity of the link connecting the browser to the server, and SSL is the latency of TCP slow-start
due to server stalling.