1,720,964 research outputs found
From Coarse Single Theoretical to Fine Multiple Practical: Machine Learning-Empowered Real-Time Communications
Get PDFL'abstract è presente nell'allegato / the abstract is in the attachmen
Where Did My Packet Go? Real-Time Prediction of Losses in Networks
Get PDFReal-time communication (RTC) platforms have undergone a consistent increase in popularity in recent years, and nowadays, they are fundamental for both work and leisure purposes. To ensure adequate Quality of Experience (QoE) for users of RTC services, we need proper traffic management policies, that, when critical network conditions are detected, react by operating either at the network configuration level or on the application to improve QoE. However, predicting critical network conditions, especially packet losses that are particularly harmful to QoE, is a very challenging task. In this paper, we propose a system for predicting packet losses that might occur in the near future (i.e., in a second) for RTP streaming traffic. We analyze several ML algorithms, from standard techniques to deep neural networks and anomaly detection algorithms, and we apply them to more than 66 hours of data from two popular RTC applications. The selection of the algorithm and its tuning turn out to be fundamental to achieving good performance. In one of the best settings, which are based on a Balanced Random Forest classifier, we obtain a recall of 0.82
Towards the Detection of Unobservable Losses in Real-Time Communications
Get PDFPacket loss, an omnipresent issue that degrades the QoE in Real-time Transport Protocol (RTP)-based real-time communications (RTC) applications, serves as a pivotal indicator for gauging network performance. Conventionally, loss detection hinges on sequence number irregularities. However, many contemporary applications incorporate customized mechanisms that diverge from the standard, confounding loss identification. Although the actual losses are transparent to applications themselves, they remain unobservable to other entities such as network operators, hampering the prospect of overall network management and performance optimization. To address this challenge, we investigate multitudinous RTC traffic gathered across various locations and times. Consequently, we uncover two types of anomalous patterns pertaining to sequence numbers. To discern between factual losses and aberrations in RTP flows, i.e., to detect the unobservable losses, we curate three distinct datasets, aggregating packets into time bins and calculating multiple traffic statistics. Subsequently, we leverage Machine Learning (ML) technologies, training the algorithm on one dataset while testing the remaining two, to classify the loss presence in a bin. Despite the inherent hurdles posed by class imbalance and intricate traffic dynamics, we achieve decent outcomes (0.64 F1-score), effectively identifying the majority of lossy bins (0.64 recall) while guaranteeing the performance for lossless scenarios (0.94 recall)
One Is Enough: Efficient Modeling of RTP Traffic for QoS Predictions in Real-Time Communications
Get PDFIn recent years, we have witnessed an unprecedented upsurge in popularity and advancement of Real-time Transport Protocol (RTP)-based real-time communication (RTC) applications. For the sake of their optimizations, Quality of Service (QoS) prediction serves as a viable venue for enhancing network monitoring and enabling preemptive solutions. However, existing methodologies are typically tailored and constrained to individual traffic flows and QoS metrics, lagging in correlation capturing and computational efficiency. In light of this, we argue that "one model is enough" to conquer these challenges, and propose a novel deep learning (DL) framework namely Oh, employing a teacher-student scheme with two training stages. The first (teacher) involves a sophisticated Long Short-Term Memory (LSTM) neural network (NN) empowered by a customized attention structure, and the second (student) comprises simple feedforward NNs to distill knowledge and reduce complexity. Specifically, Oh leverages a multi-task learning paradigm, mapping extracted features to four key QoS indicators. It is capable of simultaneously handling unlimited amount of concurrent RTP flows with packet-level information and performing end-to-end predictions of multiple QoS metrics in one single shot. Our work is based on massive traffic collected during real video-teleconferencing calls using various software, and benchmarked against multiple other machine learning (ML)/DL algorithms. As a result, Oh-teacher yields superior prediction performance, whereas Oh-student achieves distinctly enhanced temporal efficiency with comparable forecasting outcomes
Throughput Prediction in Real-Time Communications: Spotlight on Traffic Extremes
Get PDFAmidst the thriving advancement of networks, further catalyzed by the COVID-19 pandemic, we have witnessed a marked escalation in the worldwide adoption of Real-Time Communications (RTC) applications. In this context, there is a compelling necessity to cultivate intelligent and robust network infrastructures and technologies. Real-time throughput prediction emerges as a promising candidate for this purpose to foster network observability and provide preemptive functions, supporting advanced system management, e.g., bandwidth allocation and adaptive streaming. Nonetheless, contemporary solutions grapple with predicting extreme conditions in traffic throughput, notably peaks, valleys, and abrupt changes. To address the challenges, we propose a Transformer-based Deep Learning (DL) Neural Network (NN), leveraging solely packet-level information and adopting a multi-task learning paradigm, to predict short-term throughput, with an emphasis on critical values. In particular, our work is grounded in voluminous traffic traces procured from real video-teleconferencing sessions, and we formulate a time-series regression problem, comparing numerous technologies, from an adaptive filter to Machine Learning (ML) and DL approaches. Conclusively, our methodology exhibits superior efficacy, especially in forecasting traffic extremities
Modelling Concurrent RTP Flows for End-to-end Predictions of QoS in Real Time Communications
Get PDFThe Real-time Transport Protocol (RTP)-based real-time communications (RTC) applications, exemplified by video conferencing, have experienced an unparalleled surge in popularity and development in recent years. In pursuit of optimizing their performance, the prediction of Quality of Service (QoS) metrics emerges as a pivotal endeavor, bolstering network monitoring and proactive solutions. However, contemporary approaches are confined to individual RTP flows and metrics, falling short in relationship capture and computational efficiency. To this end, we propose Packet-to-Prediction (P2P), a novel deep learning (DL) framework that hinges on raw packets to simultaneously process concurrent RTP flows and perform end-to-end prediction of multiple QoS metrics. Specifically, we implement a streamlined architecture, namely length-free Transformer with cross and neighbourhood attention, capable of handling an unlimited number of RTP flows, and employ a multi-task learning paradigm to forecast four key metrics in a single shot. Our work is based on extensive traffic collected during real video calls, and conclusively, P2P excels comparative models in both prediction performance and temporal efficiency
DeX: Deep learning-based throughput prediction for real-time communications with emphasis on traffic eXtremes
Get PDFRecent years have witnessed a remarkable upsurge in the global proliferation of Real-Time Communications (RTC) applications, a trend propelled by the flourishing advancement of network technologies and further amplified by the COVID-19 pandemic. Within this context, there is a burgeoning interest in the innovation of sophisticated and intelligent network infrastructures and technologies. Positioned as a promising candidate for this purpose, real-time throughput prediction emerges as a key enabler to foster network observability and offer proactive functions, upholding advanced system management, including but not limited to, bandwidth allocation and adaptive streaming. Nonetheless, existing methodologies struggle with predicting extreme conditions of throughput, notably peaks, valleys, and abrupt changes, that are critical in RTC traffic. To surmount these obstacles, we introduce DeX, a Deep Learning (DL)-based framework, designed to predict short-term throughput, with a dexterous proficiency and dedicated focus on navigating the complexities of traffic eXtremes.
In particular, DeX leverages solely packet-level information as features and is composed of three integral components: a packet selection module that opts for an optimal subset of input features, a feature extraction
block that partially incorporates the Transformer architecture, and a multi-task learning pipeline that improves the proficiency in handling traffic extremes. Moreover, our work is anchored in extensive traffic traces garnered during actual video-teleconferencing calls, and we formulate a time-series regression problem, rigorously evaluating a spectrum of technologies ranging from an adaptive filter to diverse Machine Learning (ML) and DL approaches. Initially, we aim at predicting throughput within 500-ms time windows using historical 1024 packets out of 2048, and consequently, our methodology exhibits exceptional efficacy, especially in forecasting traffic extremities. Conclusively, we conduct a series of ablation experiments and thorough analyses to showcase the enhanced performance of various scenarios, further validating the effectiveness and robustness of DeX
BitFormer: Transformer-Based Neural Network for Bitrate Prediction in Real-Time Communications
Get PDFIn recent years, an exponential upsurge in the global proliferation of Real-Time Communications (RTC) applications has been witnessed, due to the prosperous development of networks and further fueled by the ramifications of the COVID-19 pandemic. Consequently, the imperative for development of intelligent, resilient, and scalable network infrastructures and technologies has grown significantly. Real-time bitrate prediction could play a crucial role, offering network observability and bolstering proactive system management. By accurately forecasting bitrate, it becomes possible to implement improvements at either application level or network level, such as swift and appropriate bandwidth adaptation. In this paper, we propose a novel Transformer-based deep learning framework called BitFormer designed to predict the short-term bitrate. Our work is based on extensive traffic data collected under various conditions using two prevalent RTC applications, and our model relies solely on packet-level information, which contains the fundamental traffic characteristics and facilitates effortless feature extraction. Through comprehensive evaluations and comparisons, we achieve a superior accuracy of 74% in identifying peak bitrates, while simultaneously ensuring commendable overall performance
Packet Loss in Real-Time Communications: Can ML Tame its Unpredictable Nature?
Get PDFDue to the flourishing development of networks, and abetted by the Covid-19 pandemic, we have witnessed an exponential surge in the global proliferation of Real-Time Communications (RTC) applications in recent years. In light of this, the necessity for robust, scalable, and intelligent network infrastructures and technologies has become increasingly apparent. Among the principal challenges encountered in RTC lies the issue of packet loss. Indeed, the occurrence of losses leads to communication degradation and reallocation that adversely affect the Quality of Experience (QoE). In this paper, we investigate the feasibility of predicting packet loss phenomena through the utilization of machine learning techniques, solely based on statistics derived directly from packets. We provide different definitions of packet loss, subsequently focusing on the most critical scenario, which is defined as the first loss of a series. By delineating the concept of loss, we propose different problem formulations to determine whether there exists a mathematically advantageous scenario over others. To substantiate our analysis, we demonstrate that these phenomena can be correctly identified with a recall up to 66%, leveraging three ample datasets of RTC traffic, which were collected under distinct conditions at different times, further solidifying the validity of our findings
- …
