Quality of Experience for Mobile Video Using the Smart Edge

CTN Issue: September 2017

Video is still one of the drivers of mobile bandwidth and is also a killer app for 5G success. This month Ning Wang and Chang Ge from the 5G Innovation Centre (5GIC) at the University of Surrey explore how to ensure the quality of experience needed for 4K Ultra HD video, including Virtual (VR) and Augmented Reality (AR) applications, for which the content delivery may not be well-controlled over the public internet. Wang and Ge suggest that the use of the mobile edge can make up for the variations in the backbone delivery. We welcome your comments as always.

Alan Gatherer, Editor-in-Chief

Achieving QoE-Assured 4K Mobile Video Delivery Over the Public Internet – The Role of Transport Layer

Ning Wang, Reader (Associate Professor), 5G Innovation Centre, University of Surrey
Chang Ge, Postdoctoral Researcher, 5G Innovation Centre, University of Surrey

Ning Wang
Ning Wang
Change Ge
Chang Ge

In the context of the current 5G networking, it has been a general vision that the delivery of 4K ultra-high definition (UHD) video and other immersive content applications with assured user quality of experiences (QoE) will become a representative use case in the 5G enhanced mobile broadband (eMBB) scenario. But here’s an interesting question: even with envisaged downlink capacity in 5G, e.g. >1Gbps, can a mobile user enjoy disruption-free consumption of 4K video content or other bandwidth demanding virtual and augmented reality (VR/AR) applications, regardless of where the content source is physically located in the public Internet?  Before discussing specific technical challenges associated with such a scenario, there are some common, yet noteworthy, questions to consider:  

Question 1: Do we really require 4K video quality on our smart phones with limited screen sizes?
In the case of conventional video content, there will not be any user-perceivable difference between 4K and High Definition (HD) qualities on a phone-sized screen. Nevertheless, when one is trying to consume more immersive VR/AR content using a Head Mounted Device (HMD), lower-than-4K quality will possibly make a user feel dizzy or even sick, which can be much more severe than the “mildly” poor user QoE on the traditional media.

Question 2: Haven’t the existing content delivery network (CDN) infrastructures been able to facilitate 4K video delivery already?
Yes, and no. On one hand, those subscribers to premium content services (e.g. Netflix) have been able to enjoy statistically assured 4K video quality thanks to CDN-based content distributions. However, the simple fact is that there is plenty of other content in the public Internet which is not distributed by CDNs. In this case, user-perceived video quality can never be assured based on the state-of-the-art solutions, especially if the client is far away from the actual content source. This is mainly due to the poor support from the underlying transport layer protocols across the public Internet.

Question 3: Have you heard of Google BBR (bottleneck bandwidth and round-trip propagation time? [1])
Good question! Since its introduction in late 2016, this new congestion control technique has really impressed both the research and the ordinary user communities with superb end-to-end performances compared to its predecessors, which include CUBIC and New Reno. Mobile broadband applications, including video streaming, have already seen the benefits of BBR in improving user experiences. Nevertheless, most recent study [2] has revealed some key issues with BBR, especially fairness across concurrent flow sessions which may affect its wide deployment across the whole Internet.

Coming back to the technical challenges in end-to-end mobile content delivery, typically across heterogenous data path segments (fixed backbone Internet with high bandwidth-delay-product (BDP) concatenated with radio access networks with low BDP), one major issue is that the current transport layer protocol suffers from such BDP mismatch between fixed and wireless parts of the content delivery path. This leads to very poor end-to-end throughput performance, even though adequate radio/network resources along the whole path are available. In our recently published work [3] carried out at the 5G Innovation Centre hosted by the University of Surrey, we consider the popular Dynamic Adaptive Streaming over HTTP (DASH) and HTTP Live Streaming (HLS ) based content applications, in which we tried to exploit such BDP mismatch by allowing content segments to be adaptively prefetched to the mobile edge through multiple TCP connections along the path segment with high BDP. Such adaptive prefetching operations require necessary knowledge learned by the mobile edge, including the instantaneous user QoE (e.g., video buffer status) and backhaul condition (e.g., latency). As a result, the content requests from the user device side can be directly served by the available segments that have arrived at the mobile edge just in time. Such technique can be supported by the commonly investigated edge computing paradigms [4][5] at present. In this case, there is no need to preload or cache the whole content at the mobile edge, and this is indeed a promising solution to deliver high-quality content not cached by CDNs.

One unique feature of the scheme is the virtualisation of mobile edge resources that can be leased to external service/content providers. As suggested in [3], such virtual resources can be leased to content providers who can execute content prefetching intelligence directly from the mobile edge according to dynamic conditions.  Compared to the CDN model where content objects are passively delegated to the CDN intermediaries, this approach offers specific flexibility to content providers who will be able to deploy their own content policies and intelligence with extended service capability close to end users.

According to the evaluation results obtained from real experiments based on the LTE-A platform, our scheme enables smooth playback experience at global scale through prefetching nearly 100% of all requested video segments in time at the mobile edge, which enables smooth video delivery to user devices—depending on adequate radio access network resources. In contrast, with the existing end-to-end approach, it is simply a mission impossible if a content source is beyond 100ms to reach, even with end-to-end adequate availability of bandwidth resources.

In conclusion, the key significance of our work is not necessarily confined by whether one particular scheme is able to support 4K mobile video delivery across the public Internet. Rather, it introduces a new method that can exploit the mismatched BDP characteristics across different segments of the end-to-end delivery path. On one hand, there is no doubt that the 5G-oriented radio access technologies will facilitate significant increase of user experienced data rate by hundreds of times (maybe even a thousand) compared to 4G networks.  On the other hand, more advanced media services, such as 8K, VR/AR and 6-degree of freedom (6DoF) video, will be on their way to the market, posing significantly higher data rate requirements even towards the order of gigabits per second [6]. While the race between content application requirements and network capacities will continue even beyond 5G, our vision is that the end-to-end transport protocols should never become the actual bottleneck in the future.

References

  1. Neal Cardwell, Yuchung Cheng, C. Stephen Gunn, Soheil Hassas Yeganeh, Van Jacobson, “BBR: Congestion-Based Congestion Control”, ACM Queue, Vol. 14, Issue 5, December 2016
  2. Mario Hock, Roland Bless, Martina Zitterbart, “Experimental Evaluation of BBR Congestion Control”, Proc. IEEE ICNP 2017
  3. Chang Ge, Ning Wang, Gerry Foster, M. Wilson, “Towards QoE-assured 4K Video-on-demand Delivery through Mobile Edge Virtualization with Adaptive Prefetching”, IEEE Transactions on Multimedia, Vol. 19, Issue 10, December 2017, pp. 2222-2237
  4. ETSI Multi-Access Edge Computing, http://www.etsi.org/technologies-clusters/technologies/multi-access-edge-computing
  5. Shanhe Yi, Cheng Li, Qun Li, “A Survey of Fog Computing: Concepts, Applications and Issues”, Proc. ACM Workshop on Mobile Big Data, 2015
  6. “Augmented and Virtual Reality: the First Wave of 5G Killer Apps”, Qualcomm white paper, February 2017, https://www.qualcomm.com/media/documents/files/augmented-and-virtual-reality-the-first-wave-of-5g-killer-apps.pdf

Leave a comment

Statements and opinions given in a work published by the IEEE or the IEEE Communications Society are the expressions of the author(s). Responsibility for the content of published articles rests upon the authors(s), not IEEE nor the IEEE Communications Society.