MESA: Farsighted Flow Management for Video Delivery in Wireless Networks

Video traffic over cellular and broadband wireless access networks has been rapidly increasing in the recent years. At the same time, the wireless network speeds are not increasing as quickly, thereby often leading to congestion at the wireless access links. Further, this scenario is unlikely to change significantly in the near future. In minimizing the effect of this mismatch at such bottleneck links, we explore a farsighted flow management framework, called MESA, which attempts to maximize the long-term quality of experience for each user. MESA has three key differentiating features: (1) It enables joint soft admission control and scheduling across video  flows, which we demonstrate to be a more appropriate approach for non-elastic traffic such as video, (2) it takes into account a long-term dissatisfaction metric in making admission control and scheduling decisions, which results in a uniform quality of experience to users, and (3) it builds on the resource virtualization paradigm to enable extensibility in response to future trends and innovations. We implement a prototype system of MESA and evaluate it in detail on a WiMAX testbed using both static and mobile client scenarios.

Click here for the related paper that appeared at IEEE COMSNETS 2011


The following demos show the efficacy of MESA on a WiMAX testbed.


UDP DEMO (Play with Windows Media Player or VLC player)

The following three videos compare the user perceived quality of downlink video transmission with and without MESA for RTP-based streaming.
  1. Vanilla: This case emulates traditional admission control without considering dynamic variations after flow admission. Three flows are admitted into the system with cumulative average rate below the base station capacity. However, due to fluctuations in video traffic and fluctuating wireless channel capacity, each flow perceives degraded video quality multiple times as the video progress. Notice the occasional stalls and blotchy frames.
  2. MESA-unmarked: In this case, MESA selects one flow and deprioritizes it over the other two flows. However, in this case, the source does not mark packets indicating which  packets can be dropped. User 1 gets selected in this case to be victimized. The other two users receive good quality video.
  3. MESA-marked: In this case, the source marks packets explicitly indicating which packets can be dropped in the victim slice. MESA drops these unimportant packets and improves the quality of the victimized flow too. User 2 is selected to be victimized in this case. 



TCP DEMO (Play with Windows Media Player or VLC player)

The following videos compare the user perceived quality of downlink video transmission with and without MESA for Youtube TCP based Videos.
  1. Vanilla: This case emulates traditional admission control without considering dynamic variations after flow admission. Six flows are admitted into the system with cumulative average rate below the base station capacity. However, due to fluctuations in video traffic and fluctuating wireless channel capacity, each flow perceives degraded video quality multiple times as the video progress. Notice the periodic stalls in the videos.
  2. MESA: In this case, MESA selects one or two flows and deprioritizes it over the other flows. This ensures that other clients see good quality.