Current networking solutions are unable to satisfy the low-latency requirements of real-time volumetric video conferencing when faced with heavy congestion scenarios. Traditional congestion controllers use packet loss or the change in round-trip time (RTT) to estimate the bandwidth. Commonly, this method is too slow as congestion has already occurred and the receiving user has already experienced a latency spike. Low latency, low loss and scalable throughput (L4S), recently published as RFC 9330, wants to alleviate this problem by aiming for sub 1 ms queuing delay for low-latency traffic by using accurate explicit congestion notification (AccECN) packet marking to notify applications of early congestion. We propose an L4S-based pipeline for volumetric video delivery, which achieves a more consistent latency under congestion compared to web real-time communication (WebRTC). In addition, L4S bandwidth estimation achieves a 45% faster convergence compared to Google congestion control (GCC) estimation, commonly used in WebRTC. Furthermore, in our detailed evaluation setup the L4S application experiences no packet loss, while the WebRTC-based version suffers from irrecoverable packet loss, resulting in 3% of frames being undecodable.

Nowadays, many services are offered via the cloud, i.e., they rely on interacting software components that can run on a set of connected Commercial Off-The-Shelf (COTS) servers sitting in data centers. As the demand for any particular service evolves over time, the computational resources associated with the service must be scaled accordingly while keeping the Key Performance Indicators (KPIs) associated with the service under control. Consequently, scaling always involves a delicate trade-off between using the cloud resources and complying with the KPIs. In this paper, we show that a (workload-dependent) Pareto front embodies this trade-off’s limits. We identify this Pareto front for various workloads and assess the ability of several scaling algorithms to approach that Pareto front.

Cellular Vehicle-to-Everything (C-V2X) is currently at the forefront of the digital transformation of our society. By enabling vehicles to communicate with each other and with the traffic environment using cellular networks, we redefine transportation, improving road safety and transportation services, increasing the efficiency of vehicular traffic flows, and reducing environmental impact. To effectively facilitate the provisioning of Cellular Vehicular-to-Network (C-V2N) services, we tackle the interdependent problems of service task placement and scaling of edge resources. Specifically, we formulate the joint problem and prove that it is not computationally tractable. To address its complexity we propose dhpg, a new Deep Reinforcement Learning (DRL) approach that operates in hybrid action spaces, enabling holistic decision-making and enhancing overall performance. We evaluated the performance of DHPG using simulations with a real-world C-V2N traffic dataset, comparing it to several state-of-the-art (SoA) solutions. DHPG outperforms these solutions, guaranteeing the 99th percentile of C-V2N service delay target, while simultaneously optimizing the utilization of computing resources. Finally, time complexity analysis is conducted to verify that the proposed approach can support real-time C-V2N services.

Cloud and edge Data-center (DC) are designed to allocate computing resources dynamically to users based on the agreed Service Level Agreement (SLA). However, the ever-increasing demand for beyond 5G services necessitates an efficient workload management. A key challenge in this regard is auto-scaling, a dynamic process that adjusts computing resources to meet fluctuating system demands, optimizing resource utilization and cost efficiency. Traditional auto-scaling algorithms, which rely on fixed thresholds or control-theory, may face limitations in modern DC which are characterized by diverse, dynamic, and multi-user workloads. In this paper, we propose a Reinforcement Learning (RL)-based controller that extends the capacity of the state-of-the-art RL-based auto-scalers to the multi-user workload scenario. We compare the proposed RL agent against the well-known Proportional-Integral (PI) controller and a Threshold (THD)-based controller in a multi-user workload scenario in terms of created Cloud-native Network Functions (CNFs) and peak latency performed in a discrete event simulator.

Next-generation mobile networks are expected to flaunt highly (if not fully) automated management. To achieve such a vision, Artificial Intelligence (AI) and Machine Learning (ML) techniques will be key enablers to craft the required intelligence for networking, i.e., Network Intelligence (NI), empowering myriad of orchestrators and controllers across network domains. In this paper, we elaborate on the DAEMON architectural model, which proposes introducing a NI Orchestration layer for the effective end-to-end coordination of NI instances deployed across the whole mobile network infrastructure. Specifically, we first outline requirements and specifications for NI design that stem from data management, control timescales, and network technology characteristics. Then, we build on such analysis to derive initial principles for the design of the NI Orchestration layer, focusing on (i) proposals for the interaction loop between NI instances and the NI Orchestrator, and (ii) a unified representation of NI algorithms based on an extended MAPE-K model. Our work contributes to the definition of the interfaces and operation of a NI Orchestration layer that foster a native integration of NI in mobile network architectures.

The increased popularity of social media and mobile devices has radically changed the way people consume multimedia content online. As an example, users can experience the same event (e.g. a sports event or a concert) together using social media, even if they are not in the same physical location. Moreover, the introduction of the HTTP Adaptive Streaming principle has made it possible to deliver video over the best-effort Internet with consistent quality, even for mobile devices. One of the challenges within this context is the synchronization of multimedia playback among geographically distributed clients. To solve this issue, we propose a Web-based framework which allows to synchronize the playback of different clients. We also present a novel hybrid approach for adaptive streaming to allow fast synchronization among different clients, which relies on HTTP/2's server push feature in combination with sub-second video segments. In this paper, we detail the proposed framework and provide a comprehensive analysis of its performance. Experiments show that the novel hybrid approach can reduce synchronization time with 19.4% compared to standard adaptive streaming over HTTP/1.1 when bandwidth is limited to 2.5 Mb/s and an RTT of 150 ms. The gain increases even more when a higher throughput is available. The obtained results entail that the proposed framework can provide quality of experience for all users watching online video together.

Video streaming services have continuously gained popularity over the last decades, accounting for about 70% of all consumer Internet traffic in 2016. All of these video streaming sessions have strict delivery deadlines in order to avoid playout interruptions, detrimentally impacting the Quality of Experience (QoE). However, the vast majority of this traffic uses TCP at the transport layer, which is known to be far from minimizing the number of deadline-missing streams. By introducing deadline-awareness at the transport layer, video delivery can be optimized by prioritizing specific flows. This paper proposes a deadline-aware congestion control mechanism, based on a parametrization of the traditional TCP New Reno congestion control strategy. By taking into account the available deadline information, the modulation of the congestion window is dynamically adapted to steer the aggressiveness of a considered stream. The proposed approach has been thoroughly evaluated in both a video-on-demand (VoD)-only scenario and a scenario where VoD streams co-exist with live streaming sessions and non-deadline-aware traffic. It was shown that in a video streaming scenario the minimal bottleneck bandwidth can be reduced by 16% on average when using deadline-aware congestion control. In coexistence with other TCP traffic, a bottleneck reduction of 11% could be achieved.

Recently, video-on-demand (VoD) streaming services like Netflix and Hulu have gained a lot of popularity. This has led to a strong increase in bandwidth capacity requirements in the network. To reduce this network load, the design of appropriate caching strategies is of utmost importance. Based on the fact that, typically, a video stream is temporally segmented into smaller chunks that can be accessed and decoded independently, cache replacement strategies have been developed that take advantage of this temporal structure in the video. In this paper, two caching strategies are proposed that additionally take advantage of the phenomenon of binge watching, where users stream multiple consecutive episodes of the same series, reported by recent user behavior studies to become the everyday behavior. Taking into account this information allows us to predict future segment requests, even before the video playout has started. Two strategies are proposed, both with a different level of coordination between the caches in the network. Using a VoD request trace based on binge watching user characteristics, the presented algorithms have been thoroughly evaluated in multiple network topologies with different characteristics, showing their general applicability. It was shown that in a realistic scenario, the proposed election-based caching strategy can outperform the state-of-the-art by 20% in terms of cache hit ratio while using 4% less network bandwidth.

The growing popularity of over the top (OTT) video streaming services has led to a strong increase in bandwidth capacity requirements in the network. By deploying intermediary caches, closer to the end-users, popular content can be served faster and without increasing backbone traffic. Designing an appropriate replacement strategy for such caching networks is of utmost importance to achieve high caching efficiency and reduce the network load. Typically, a video stream is temporally segmented into smaller chunks that can be accessed and decoded independently. This temporal segmentation leads to a strong relationship between consecutive segments of the same video. Therefore, caching strategies have been developed, taking into account the temporal structure of the video. In this paper, we propose a novel caching strategy that takes advantage of clients announcing which videos will be watched in the near future, e.g., based on predicted requests for subsequent episodes of the same TV show. Based on a Video-on-Demand (VoD) production request trace, the presented algorithm is evaluated for a wide range of user behavior and request announcement models. In a realistic scenario, a performance increase of 11% can be achieved in terms of hit ratio, compared to the state-of-the-art.

In this paper we analyze a trace of a deployed VoD (video on demand) system. Users issue requests for content items in an online VoD catalogue at given moments in time. Based on this information alone we identify communities of users that have similar content preferences, which we refer to as implicit social communities. We find that there is evidence for a limited number of groups of similar users. Next we also determine lead users, i.e., users that consume popular content items consistently before other users do. We show that such users can be identified in the considered data set. We also explain how these two pieces of information could be used to improve recommendation systems and content distribution networks.

Video streaming, in particular, hypertext transfer protocol based (HTTP) adaptive streaming (HAS) of video, is expected to be a dominant application over mobile networks in the near future. The observation that the base station can alter the video quality requested by a HAS client to its server by controlling the over-the-air throughput from the base station to the client implies that the base station can jointly maximize aggregate video quality of all the HAS flows and throughput of data flows that it serves. We formulate a utility maximization problem that separately takes into account different utility functions for video and data flows and show that the utility maximization can be achieved through an algorithm, we term adaptive guaranteed bit rate (AGBR), wherein target bit rates are calculated for each HAS flow and passed on to an underlying minimum rate proportional fair scheduler that schedules resources across all the flows. This approach has the advantage that it retains the existing scheduling function in the base station with a separate function to compute the target bit rates for the video flows allowing them to only change slowly over time in order to avoid frequent video quality changes. Through analytical modeling and simulations we show that the proposed algorithm can achieve required fairness among the video flows as well as automatically and fairly adapt video quality with increasing congestion thereby preventing data flow throughput starvation.

In this paper, we investigate the proxy video caching problem and propose a specific caching replacement strategy for segmented video streaming. The proposed caching algorithm makes use of the natural linear time structure of video streaming, which fits the philosophy of the optimal MIN algorithm. By observing the watching positions of active viewers at a certain point of time, the replacement approach computes the exact reuse time or, if not available, the predicted reuse time for stored video segments, and discards the segment that is supposedly to be used the furthest in time. Evaluations performed on a trace of a real Video-On-Demand (VOD) service show that the proposed caching algorithm substantially improves the hit ratio, especially for small caches.

HTTP-based delivery for Video on Demand (VoD) has been gaining popularity within recent years. Progressive Download over HTTP, typically used in VoD, takes advantage of the widely deployed network caches to release video servers from sending the same content to a high number of users in the same VoD service. However, due to the inherent heterogeneity of user demands, which may result in requesting the same video content in different resolutions or qualities, the caching efficiency is expected to decrease due to a higher variety in requested media files. The use of Scalable Video Coding allows different representations of the same content to be combined in a single file, whose parts, aka layers, are requested sequentially by a user up to the maximum desired quality. In this paper we show the benefits of using Scalable Video Coding to maintain the same set of possible video content representations, while at the same time maximizing the caching efficiency.

HTTP-based delivery for Video on Demand (VoD) has been gaining popularity within recent years. Progressive Download over HTTP, typically used in VoD, takes advantage of the widely deployed network caches to relieve video servers from sending the same content to a high number of users in the same access network. However, due to a sharp increase in the requests at peak hours or due to cross-traffic within the network, congestion may arise in the cache feeder link or access link respectively. Since the connection characteristics may vary over the time, with Dynamic Adaptive Streaming over HTTP (DASH), a technique that has been recently proposed, video clients may dynamically adapt the requested video quality for ongoing video flows, to match their current download rate as good as possible. In this work we show the benefits of using the Scalable Video Coding (SVC) for such a DASH environment.

We consider a centralised (client-server) digital TV network with heterogeneous receiver devices of different resolutions, requiring a multi-rate transport system. There exist two main ways to store and transport (streamed) TV channels in such a system: either by providing different single-layer versions of a channel (simulcast transport mode) or by keeping one multi-layered version (encoded e.g. in SVC) with extractable substreams. We propose one approximate analytical and two simulation methods to estimate the capacity demand in such a network with variable bit rate channels and we consider two behaviour models. In some TV distribution networks, the video is delivered in constant bit rate. However, this implies that the video quality is varying. In order to provide better quality of service (QoS), a network operator must deliver the channels in non-constant bit rate aiming in this way at constant video quality. Our models take into account also the correlations between the different resolutions of a channel. Starting from real experimental data, we obtain the necessary input to our models and explore two realistic TV network scenarios - with bouquets of 50 and 300 channels, respectively. The results by the three approaches correspond well (relative error of 0.5% at most). In the case of 50 channels, SVC outperforms simulcast in terms of required bandwidth, while in the case of 300 channels, SVC is outperformed by simulcast. Therefore, we conclude that it depends on the system parameters which of both transport strategies will be more beneficial to save network resources.

This paper presents three approaches to estimate the required resources in an infrastructure where digital TV channels can be delivered in unicast or multicast (broadcast) mode. Such situations arise for example in Cable TV, IPTV distribution networks or in (future) hybrid mobile TV networks. The three approaches presented are an exact calculation, a Gaussian approximation and a simulation tool. We investigate two scenarios that allow saving bandwidth resources. In a static scenario, the most popular channels are multicast and the less popular channels rely on unicast. In a dynamic scenario, the list of multicast channels is dynamic and governed by the users' behavior. We prove that the dynamic scenario always outperforms the static scenario. We demonstrate the robustness, versatility and the limits of our three approaches. The exact calculation application is limited because it is computationally expensive for cases with large numbers of users and channels, while the Gaussian approximation is good exactly for such systems. The simulation tool takes long to yield results for small blocking probabilities. We explore the capacity gain regions under varying model parameters. Finally, we illustrate our methods by discussing some realistic network scenarios using channel popularities based on measurement data as much as possible.

In First Person Shooter (FPS) games the Round Trip Time (RTT), i.e., the sum of the network delay from client to server and the network delay from server to client, impacts the gamer's performance considerably. Game client software usually has a built-in process to measure this RTT (also referred to as ping time), and therefore gamers do not want to connect to servers with a long ping time. This paper develops a methodology to evaluate the ping time in a scenario where gamers access a common gaming server over an access network, consisting of a link per user that connects this user to a shared aggregation node that in turn is connected to the gaming server via a bottleneck link. First, a model for the traffic the users and the server generate, is proposed based on experimental results of previous papers. It turns out that the characteristics of the (downstream) traffic from server to clients differ substantially from the characteristics of the client-to-server (upstream) traffic. Then, two queuing models are developed (one for the upstream and one for the downstream direction) and combined such that a quantile of the RTT can be calculated given all traffic and network parameters (packet sizes, packet inter-arrival times, link rate, network load, ...). This methodology is subsequently used to assess the (quantile of the) RTT in a typical Digital Subscriber Line (DSL) access scenario. In particular, given the capacity dedicated to gaming traffic on the bottleneck link (between the aggregation node and gaming server), the number of gamers (or equivalently the gaming load the bottleneck link can support) is determined under the restriction that the quantile of the RTT should not exceed a predefined bound. It turns out that this tolerable load is surprisingly low in most circumstances. Finally, it is remarked that this conclusion depends to some extent on the details of the downstream traffic characteristics and that measurements reported in literature do not give conclusive evidence on the exact value of all parameters, such that, although the qualitative conclusion still remains valid, additional experiments could refine the detailed quantitative results.

This paper analyzes the performance of a large population composed of several classes of long lived TCP flows experiencing packet losses due to random transmission errors and to congestion created by the sharing of a common tail-drop or RED bottleneck router. Each class has a different transmission error rate. This setting is used to analyze the competition between wired and wireless users in an access network, where one class (the wired class) has no or small (like BER in DSL) transmission error losses whereas the other class has higher transmission error losses, or the competition between DSL flows using different coding schemes. We propose a natural and simple model for the joint throughput evolution of several classes of TCP flows under such a mix of losses. Two types of random transmission error losses are considered: one where losses are Poisson and independent of the rate of the flow, and one where the losses are still Poisson but with an intensity that is proportional to the rate of the source. We show that the large population model where the population tends to infinity has a threshold (given in closed form) below which there are no congestion losses at all in steady state, and above which there is a stationary limiting regime in which we can compute both the mean value and the distribution of the rate obtained by each class of flow. We also show that the maximum mean value for the aggregated rate is achieved at the threshold.

In this paper, we consider a set of HTTP flows using TCP over a common drop-tail link to download files. After each download, a flow waits for a random think time before requesting the download of another file, whose size is also random. When a flow is active its throughput is increasing with time according to the additive increase rule, but if it suffers losses created when the total transmission rate of the flows exceeds the link rate, its transmission rate is decreased. The throughput obtained by a flow, and the consecutive time to download one file are then given as the consequence of the interaction of all the flows through their total transmission rate and the link's behavior.We study the mean-field model obtained by letting the number of flows go to infinity. This mean-field limit may have two stable regimes : one without congestion in the link, in which the density of transmission rate can be explicitly described, the other one with periodic congestion epochs, where the inter-congestion time can be characterized as the solution of a fixed point equation, that we compute numerically, leading to a density of transmission rate given by as the solution of a Fredholm equation. It is shown that for certain values of the parameters (more precisely when the link capacity per user is not significantly larger than the load per user), each of these two stable regimes can be reached depending on the initial condition. This phenomenon can be seen as an analogue of turbulence in fluid dynamics: for some initial conditions, the transfers progress in a fluid and interaction-less way; for others, the connections interact and slow down because of the resulting fluctuations, which in turn perpetuates interaction forever, in spite of the fact that the load per user is less than the capacity per user. We prove that this phenomenon is present in the Tahoe case and both the numerical method that we develop and simulations suggest that it is present in the Reno case too. It translates into a bi-stability phenomenon for the finite population model within this range of parameters.

Estimating correct traffic descriptors is an important issue when deploying IP telephony with quality of service guarantees. Usually, a token bucket is used to police a traffic flow. Such a policer checks (and ensures) that the traffic flow fits inside a traffic envelope. For variable bit rate (VBR) sources normally two traffic envelopes are specified: one dealing with the peak rate and one dealing with the sustainable rate with an associated burst size. These parameters have to be specified when making a reservation with either the resource reservation protocol (RSVP) in case of IntServ or may be specified in a service level specification (SLS) when using DiffServ. This paper shows how to choose the traffic descriptors for aggregate voice traffic for both constant bit rate sources and sources that use voice activity detection. Also the influence of variable packet size on the packet discard ratio (PDR) is shown. A traditional policer discards large packets more frequent than smaller ones. To alleviate this, a modification to the token bucket algorithm is proposed.

We develop a methodology to set the VoIP application parameters (voice packet size and dejittering delay) and the UMTS air interface parameters in such a way that the quality of VoIP calls involving a UMTS party is ensured. We use analytical techniques to determine the delay and packet loss contributions of the various transmission stages crossed by the voice packet flow, and the E-model to predict the perceived quality. Our numerical results show that provided the parameters of the VoIP application and the UMTS air interface are chosen properly, UMTS access and the stringent delay and packet loss requirements of VoIP are reconcilable.

This paper discusses resource allocation and management in Differentia ted Services (DiffServ) networks, particularly in the context of IP telephony. We assume that each node uses Weighted Fair Queuing (WFQ) schedulers in order to provide Quality of Service (QoS) to aggregates of traffic. All voice traffic destined for a certain output interface is aggregated into a single queue. When a voice flow traverses a node, its packets are placed in this queue; which is drained at a certain rate, determined by a weight associated with that queue. This paper shows how to set this weight such that the edge-router-to-edge-router queuing delay in the DiffServ network is statistically bounded. The nodes are modeled as M/G/1 queuing systems and a heuristic formula is used to compute a quantile of the queuing delay. This formula is compared with results derived from simulations. It is also shown how to use this result to allocate the appropriate resources in a DiffServ network and how to update and process RSVP messages at the edge of a DiffServ network in an IntServ over DiffServ scenario.

We assess the performance of a thresholded merging technique used as part of a procedure to obtain a motion-based segmentation of an image sequence. First an initial estimate for the motion field is obtained by using an improved block matching method. Then, an intensity-based initial segmentation is performed, using watershed segmentation on a non-linearly diffused version of the image. Next, a motion vector based on the initial motion field is calculated for each segment, and, finally, the obtained segments and their motion vectors are fed into the motion-based merging scheme, yielding the final segmentation. Results for a thresholded minimum distance merging technique are given and are compared with a K-means clustering algorithm.

A private communication system for two-wire full-duplex transmission at an information rate of 144 kb/s that uses time compression multiplex (TCM) over short cables is considered. A low-cost receiver without pulse-shaping prefilter or equalizer that derives its timing from the estimated start-bit arrival times is presented and analyzed. The worst-case analysis involves the derivation of the timing error distribution and the bit error rate, along with an accurate approximation for the timing error variance. It is shown that the bit error rate degradation due to timing errors can be limited to less than 1.5 dB in signal-to-noise ratio by using a recursive start-bit detection algorithm instead of a conventional memoryless algorithm.>