Suppose we place a wireless access point (WAP) at every gas station. The Internet bandwidth for the WAP is 1 Mbps. The communication range for the WAP is 100 meters (i.e., any computer within 100 meters from a WAP can communicate with the WAP). The wireless capacity is 11 Mbps. A taxi passes by a gas station once every 10 minutes at the speed of 30 miles per hour. Our goal is to deliver (stored) streaming media to the smartphone of a passenger on the taxi. We have two options: (i) let the smartphone directly download streaming media from the Internet whenever the taxi is in contact with a WAP, (ii) predict the set of WAPs that the taxi will visit, push streaming media to WAPs before they are visited by the taxi, and let the smartphone directly download pre-fetched locally from WAPs (instead of the Internet). In both options, with sufficient buffering, the smartphone can play streaming media smoothly even when it is offline (i.e. not in contact with any WAPs). Clearly, the second option can potentially achieve much higher average data rate. The following two questions try to quantify the potential performance benefit of prefetching.

(3a) Download from Internet without prefetching. Suppose that we can download data at the full Internet/wireless capacity (i.e. we ignore loss, transport dynamics, MAC dynamics, protocol overhead, etc.) What is the average streaming media data rate that we can sustain by using Internet download only without prefetching?

(3b) Download prefetched data from WAP. Suppose that we can download data at the full Internet/wireless capacity. Also suppose we can accurately predict the set of WAPs that the taxi will visit. What is the average streaming media data rate that we can sustain by downloading prefetched data from WAPs?