Headlight prefetching for mobile media streaming

Authors:

• Shiow-yang Wu National Dong Hwa University
• Jungchu Hsu E-TEN Information Systems
• Chieh-Ming Chen Microelectronics Technology

Complete Citation

Wu, S., Hsu, J., and Chen, C. 2007. Headlight prefetching for mobile media streaming. In Proceedings of the 6th ACM international Workshop on Data Engineering For Wireless and Mobile Access (Beijing, China, June 10 - 10, 2007). MobiDE? '07. ACM, New York, NY, 67-74. DOI= http://doi.acm.org/10.1145/1254850.1254862

Abstract

Multimedia information services in mobile environments are becoming more and more important with the proliferation of 3G technologies. Media streaming, in particular, is a promising technology for providing services such as news clips, live sports, and hot movies on the fly. To avoid service interruption when the users keep moving, proper data management strategies must be employed. We propose a new headlight prefetching technique for the streaming access points to deal with the uncertainty of client movement and the requirement of seamless service hand-off. For each mobile client, we maintain a virtual fan-shaped prefetching zone along the direction of movement similar to the headlight of a moving vehicle. The overlapping area and the accumulated virtual illuminance of the headlight zone on a particular cell determines the degree and volume of prefetching to be made by the streaming access point of that cell. Headlight prefetching solves the issues of identifying the streaming access points responsible for prefetching, the timing and the amount of data to prefetch in a single mechanism which is simple and effective. Simulation results demonstrate that our techniques can significantly decrease streaming disruptions, reduce bandwidth consumption, increase cache utilization and improve service response time.

Annotations

This paper aims to increase the reliability of streaming media for a mobile client. The insperation for their scheme is how a head-light of a car projects the where the car will move to, they make the same assumptions of a mobile user. In order to increase the reception of the stream, the data is prefetched to all of the wireless routers that are on the projected path (with a cut-off) of the wireless user. Now, in order to help eliminate the horrible amount of overhead that this would place on the system, they argue that many different clients will likely access the same stream, so they can get the benefit of shared cone overlap.

The second major issue that they address is what happens is when the direction of a client abruptly changes, such that the new headlight cone shares nothing with the old. These present a few methods to resolve this issue, each with their plusses and minuses. Overall the best performance (in terms of interruption time and bandwidth consumption) is a closest neighbor approach where they calculate the closest neighbor to the old cone and use that information to determine the new cone. This is also the most computationally expensive operation.

In their results, they so improvement in the reliability of the stream by useing their various techniques. They show that if the access pattern is random they have a reduction of 75% of the number of interruptions, and if the requests follows a Zipf distribution, then the reduction is slightly over 55%. The speed of the client for these tests is taken to be 50 m/s. While they mention turning at a rate of 0 - 90 degress, they do not indicate how often they turn (which will have some impact on the results).

-- DavidSalyers - 09 Apr 2008