Mobile Data Offloading: How Much Can WiFi Deliver? Kyunghan Lee, Injong Rhee, Joohyun Lee, Song Chong, Yung Yi CoNEXT Presentor: Seokshin Son
Contents Introduction Measurement –Temporal Coverage –Inter-connection Time –End-to-end Rates Simulation –Offloading Efficiency –Completion Time –Energy Consumption Conclusion 2/20
Introduction
Quantitative study on the performance of 3G mobile data offloading through WiFi networks –Measure users’ WiFi connectivity –Trace-driven simulation Focuses on temporal offloading –Effects on delayed transfer –Delay deadline and offloading efficiency 4/20
Introduction 5/20
Measurement
Background application of iPhone –WiFi AP location –Connection time / duration –Data transfer rate Gathered data is uploaded to server daily Recorded every 3 min. 7/20
Temporal Coverage Temporal coverage > Spatial coverage 8/20
Inter-connection Time 9/20 Well fit to Weibull distribution
End-to-End Rates 10/20 Sleep with home AP
Simulation
Simulation Model DurationInterval All day2 hr.40 min. Daytime50 min.25 min. User mobility pattern is reflected Choices: Distribution of IAT / FS and value of IAT / FS 12/20
Traffic Model Average Inter Arrival Time (IAT) distribution is assumed exponential Many application generates heavy-tailed traffic [2] Traffic demand in 2014 DL: Deadline From Estimated traffic 13/20
Offloading Efficiency In [1], 10-30% gains with 100 sec. deadline –With bus or war-driving traces Tens of minutes of deadline is needed for substantial gain. (= WiFi transferred bytes / total bytes) 14/20
Completion Time Deadline > Completion time 3G rate: 200 Kbps (filesize, deadline) Completion time – Waiting time ∝ 1 / battery power saving) 15/20
Energy Consumption On-the-spot offloading already achieves 55% energy saving Delayed offloading can save substantial energy only with long deadline 16/20
Impact of Traffic Types Huge file arrives -> less efficient –System capacity also decreases 17/20
Impact of WiFi Deployment Eliminate APs according to connection time: Random and Activity-based Careful deployment plan will be useful 18/20
Conclusion
Simulation model –Reflects daily mobility patterns of average users On-the-spot offloading saves –65% of mobile user traffic –55% of battery power Allowing delays saves more traffic and battery power –As well as load balancing 20/20
References [1] A. Balasubramanian, R. Mahajan, and A. Venkataramani. Augmenting mobile 3g using wifi. In Proceedings of ACM MobiSys, [2] A. Abhari and M. Soraya.Workload generation for youtube. Multimedia Tools and Applications, 46(1):91–118, /20
Appendix: Weibull Distribution k: shape parameter λ: scale parameter 22/20