A growing trend for information technology is to not just react to changes, but anticipate them as much as possible. This paradigm made modern solutions, such as recommendation systems, a ubiquitous presence in today’s digital transactions. Anticipatory networking extends the idea to communication technologies by studying patterns and periodicity in human behavior and network dynamics to optimize network performance. This survey collects and analyzes recent papers leveraging context information to forecast the evolution of network conditions and, in turn, to improve network performance. In particular, we identify the main prediction and optimization tools adopted in this body of work and link them with objectives and constraints of the typical applications and scenarios. Finally, we consider open challenges and research directions to make anticipatory networking part of next generation networks.

/pdf/a-survey-of-anticipatory-mobile-networking-context-based-2yy8uql5nk.pdf

A Survey of Anticipatory Mobile Networking: Context-Based Classification, Prediction Methodologies, and Optimization Techniques

Recently, Hypertext Transfer Protocol (HTTP)-based adaptive streaming has become the de facto standard for video streaming over the Internet. It allows clients to dynamically adapt media characteristics to the varying network conditions to ensure a high quality of experience (QoE)—that is, minimize playback interruptions while maximizing video quality at a reasonable level of quality changes. In the case of live streaming, this task becomes particularly challenging due to the latency constraints. The challenge further increases if a client uses a wireless access network, where the throughput is subject to considerable fluctuations. Consequently, live streams often exhibit latencies of up to 20 to 30 seconds. In the present work, we introduce an adaptation algorithm for HTTP-based live streaming called LOLYPOP (short for low-latency prediction-based adaptation), which is designed to operate with a transport latency of a few seconds. To reach this goal, LOLYPOP leverages Transmission Control Protocol throughput predictions on multiple time scales, from 1 to 10 seconds, along with estimations of the relative prediction error distributions. In addition to satisfying the latency constraint, the algorithm heuristically maximizes the QoE by maximizing the average video quality as a function of the number of skipped segments and quality transitions. To select an efficient prediction method, we studied the performance of several time series prediction methods in IEEE 802.11 wireless access networks. We evaluated LOLYPOP under a large set of experimental conditions, limiting the transport latency to 3 seconds, against a state-of-the-art adaptation algorithm called FESTIVE. We observed that the average selected video representation index is by up to a factor of 3 higher than with the baseline approach. We also observed that LOLYPOP is able to reach points from a broader region in the QoE space, and thus it is better adjustable to the user profile or service provider requirements.

/pdf/qoe-based-low-delay-live-streaming-using-throughput-1tqouyioew.pdf

QoE-Based Low-Delay Live Streaming Using Throughput Predictions

Federated Learning (FL) is very appealing for its privacy benefits: essentially, a global model is trained with updates computed on mobile devices while keeping the data of users local. Standard FL infrastructures are however designed to have no energy or performance impact on mobile devices, and are therefore not suitable for applications that require frequent (online) model updates, such as news recommenders. 
This paper presents FLeet, the first Online FL system, acting as a middleware between the Android OS and the machine learning application. FLeet combines the privacy of Standard FL with the precision of online learning thanks to two core components: (i) I-Prof, a new lightweight profiler that predicts and controls the impact of learning tasks on mobile devices, and (ii) AdaSGD, a new adaptive learning algorithm that is resilient to delayed updates. 
Our extensive evaluation shows that Online FL, as implemented by FLeet, can deliver a 2.3x quality boost compared to Standard FL, while only consuming 0.036% of the battery per day. I-Prof can accurately control the impact of learning tasks by improving the prediction accuracy up to 3.6x (computation time) and up to 19x (energy). AdaSGD outperforms alternative FL approaches by 18.4% in terms of convergence speed on heterogeneous data.

/pdf/fleet-online-federated-learning-via-staleness-awareness-and-i1ym03cevs.pdf

FLeet: Online Federated Learning via Staleness Awareness and Performance Prediction

Throughput prediction is essential for ensuring high quality of service for video streaming transmissions. However, current methods are incapable of accurately predicting throughput in mobile networks, especially for moving user scenarios. Therefore, we propose a TCP throughput prediction method TRUST using machine learning for mobile networks. TRUST has two stages: user movement pattern identification and throughput prediction. In the prediction stage, the long short-term memory (LSTM) model is employed for TCP throughput prediction. TRUST takes all the communication quality factors, sensor data and scenario information into consideration. Field experiments are conducted to evaluate TRUST in various scenarios. The results indicate that TRUST can predict future throughput with higher accuracy than the conventional methods, which decreases the throughput prediction error by maximum 44% under the moving bus scenario.

TRUST: A TCP Throughput Prediction Method in Mobile Networks

To enable highly automated driving and the associated comfort services for the driver, vehicles require a reliable and constant cellular data connection. However, due to their mobility vehicles experience significant fluctuations in their connection quality in terms of bandwidth and availability. To maintain constantly high quality of service, these fluctuations need to be anticipated and predicted before they occur. To this end, different techniques such as connectivity maps and online throughput estimations exist. In this paper, we investigate the possibilities of a large-scale future deployment of such techniques by relying solely on lowcost hardware for network measurements. Therefore, we conducted a measurement campaign over three weeks in which more than 74,000 throughput estimates with correlated network quality parameters were obtained. Based on this data set—which we make publicly available to the community—we provide insights in the challenging task of network quality prediction for vehicular scenarios. More specifically, we analyse the potential of machine learning approaches for bandwidth prediction and assess their underlying assumptions.

Cellular Bandwidth Prediction for Highly Automated Driving - Evaluation of Machine Learning Approaches based on Real-World Data.

Bandwidth-sensitive applications such as adaptive video streaming rely on accurate prediction of future network throughput to enable them to react to and compensate for the rapidly fluctuating bandwidth often found in mobile networks. Researchers have developed various prediction algorithms in the literature of which many have been employed in real-world applications. However, there is a lack of systematic study on the comparative performance of the existing prediction algorithms in the context of mobile networks. This work addresses this void by conducting a systematic performance comparison of 7 prediction algorithms, and analyzes their characteristics when applied to the prediction of TCP throughput in mobile networks. The performance results are obtained from extensive trace-driven simulations where the throughput trace data were captured in production 3G/HSPA mobile networks in 3 locations over a period of 9 months and hence offer a good representation of the prediction algorithms' real- world performance. Furthermore, we applied the theory of differential entropy in information theory to obtain an estimated lower bound on throughput prediction errors which, for the first time, enables one to evaluate the absolute performance of these prediction algorithms. The results revealed that more complex algorithms are not necessarily better, and there exists a specific range of operating parameters where predictions are generally more accurate.

An Empirical Study of Throughput Prediction in Mobile Data Networks

Fueled by the growth of 3G/4G mobile networks, mobile video streaming has become one of the main applications in the mobile Internet. Due to mobile networks’ inherent bandwidth fluctuations, the industry as well as researchers have developed many adaptive streaming algorithms to compensate for such fluctuations to improve streaming performance. Given the wide range of network settings, it is not surprising that existing algorithms can and do perform differently across different network and system conditions. This work breaks away from the conventional one-size-fits-all approach to designing adaptive streaming systems by developing a new framework called PSRA where past throughput trace data - captured as a by-product of streaming, are analyzed to construct a statistical model to automatically tune the adaptation algorithm for future streaming sessions according to the underlying network and system configurations. Compared to existing approaches, the PSRA-optimized streaming algorithm can achieve predictable, consistent, and controllable streaming performance across a wide-range of network and system configurations. Moreover, PSRA offers to service provider a new tool to precisely control the tradeoff between video quality and streaming performance. Results from extensive trace-driven simulations as well as experiments verified PSRA's performance under real-world mobile network and system configurations.

Post-Streaming Rate Analysis—A New Approach to Mobile Video Streaming with Predictable Performance

Adaptive video streaming is an essential tool for improving the performance of video delivery over mobile networks. By dynamically switching between different bit-rate versions of the same video, adaptive video streaming can compensate for and adapt to the ever-changing network conditions inherent in today's 3G/4G networks. However, existing adaptive streaming algorithms, both academic and commercial ones, do not offer any prediction on the streaming performance of future streaming sessions, nor allow the content providers or users to explicitly control the tradeoff between streaming performance and video quality. This study tackles this fundamental challenge by developing a novel framework called Throughput-Differentiated-Rate-Adaptive Post-Streaming-Rate-Analysis (TDRA-PSRA) based on a new statistical model that directly relates past bandwidth statistics to future streaming performance. Extensive simulation results obtained from real-world mobile network trace data revealed three remarkable properties of TDRA-PSRA: (a) the actual average streaming performance is very close to the target set forth by the ICPs/users; (b) it enables the ICPs/users to control the tradeoff between streaming performance and video quality; (c) it offers a mean to directly control the frequency of bit-rate switches — a key factor to subjective video quality.

On adaptive video streaming with predictable streaming performance

Mobile video streaming is one of the fastest growing applications in the mobile Internet. Nevertheless, delivering high-quality streaming video over mobile networks remains a challenge. Researchers have since developed various novel streaming algorithms such as rate-adaptive streaming to improve the performance of mobile streaming services. However, selection or optimization of streaming algorithms is far from trivial and there is no systematic way to incorporate the tradeoffs between various performance metrics. This work aims at attacking the heart of the problem by developing a novel framework called Post Streaming Quality Analysis (PSQA) to automatically tune any streaming algorithms to maximize a given quality-of-experience (QoE) objective. We show that PSQA not only can be applied to optimize the performance of existing streaming algorithms, but also opens a new way for the exploration of new adaptive video streaming protocols and QoE metrics. Simulation results based on real network throughput traces show that PSQA can optimize existing and new streaming algorithms to achieve QoE that is remarkably close to the optimal achieved using brute-force method ex post facto.

A unified framework for automatic quality-of-experience optimization in mobile video streaming

Much work has been done to improve the performance of video streaming over mobile data networks. The widespread adoption of HTTP/TCP for video streaming further complicates the problem as TCP's own dynamics add even more fluctuations to the already unpredictable network bandwidth. Not surprisingly, none of the existing video bit-rate selection algorithms can achieve consistent or predictable streaming performance in the presence of wide network bandwidth fluctuations. This work tackles this problem by developing a new model to capture the statistical correlations between streaming performance, video bit-rate selection, and past TCP throughput data. Based on a novel post-streaming rate analysis technique, the proposed model can be used to predict the streaming performance of future video sessions at any given video bit-rate choices. This not only enables content providers to control the quality-of-service for streaming users, but also to ensure consistent streaming performance for users with widely different access bandwidths.

Yan Liu

Papers

An Empirical Study of Throughput Prediction in Mobile Data Networks

Post-Streaming Rate Analysis—A New Approach to Mobile Video Streaming with Predictable Performance

On adaptive video streaming with predictable streaming performance

A unified framework for automatic quality-of-experience optimization in mobile video streaming

Providing predictable streaming performance in mobile video streaming