The rapid growth of mobile-based global live streaming services has intensified the challenge of delivering high-quality, real-time video transmission, particularly due to the mismatch between diverse network requirements and rigid resource adaptation strategies. Traditional multipath transmission approaches often lack the dynamic adaptability needed to support service diversity, with isolated path selection leading to resource contention and static or coarse-grained allocation mechanisms failing to respond promptly to network changes. These limitations negatively impact both service quality and overall resource utilization. To address these challenges, this paper presents a flexible real-time service transmission system. Specifically, the system introduces a service diversity-aware mechanism coupled with collaborative optimization of network resource paths, enabling high-quality differentiated transmission services. Furthermore, a dynamic routing optimization algorithm is proposed, leveraging In-band Network Telemetry (INT) technology to monitor network resources in real time and flexibly generate optimal routing strategies. We conducted long-distance (over 2000 kilometers) cross-domain testing in real-world large-scale Information Service Providers (ISPs) networks. Experimental results demonstrate that our solution can operate effectively in existing solutions while outperforming current approaches in terms of Quality of Service (QoS).

Delay-sensitive applications like telemedicine and VR/AR intensify competition for network resources and elevate congestion risks, particularly in mobile networks with highly dynamic link conditions. Traditional end-to-end congestion control methods suffer from prolonged response times, rendering them ineffective for Delay-sensitive applications. To this end, this paper proposes a novel In-Network Congestion Control (INCC) mechanism that accelerates congestion control by enabling network nodes to proactively identify bottlenecks and promptly notify end-hosts. Unlike traditional end-host-centric approaches, INCC facilitates collaborative congestion decision-making between end-hosts and in-network unit. INCC classifies congestion into two phases: “yellow” and “red” based on the local queue length bottleneck awareness and global congestion flow bottleneck statistics. For the “yellow” local congestion phrase, we design an in-network local control algorithm that performs proactive packet dropping and rate adjustment to mitigate emerging congestion. For the “red” global congestion phrase, we design an end-host and network cooperative global congestion control algorithm to make precise sending rate adaptation by proactive bottleneck awareness. We implement INCC via Linux kernel modifications and design three experiments to compare with Cubic, NewReno, and BBR. Experimental results demonstrate INCC has good performance on round-trip time and throughput, achieving 99.03% scheduling fairness in flow contention scenarios. Additionally, INCC has low execution overhead on CPU utilization and realize microsecond computational latency.

Recently, the popularity of crowdsourced live streaming (CLS) has increased significantly. To overcome the bandwidth limitations of the broadcaster’s stream in the first mile, some research has introduced video super-resolution (VSR) algorithms at the source server. However, this implementation brings an additional computational burden. In VSR-enabled CLS, the stream push bitrate of the broadcaster will directly determine the volume of data transmitted in the first mile and processed by the source server. An inappropriate upload bitrate can lead to unacceptable delays and resource wastage. To address this issue, this paper proposes a VSR-UAiC framework, which implements an upload adaptive bitrate (ABR) control in VSR-enabled CLS. The VSR-UAiC framework is capable of perceiving multi-dimensional information, including network conditions, computing power, video content, and viewer requests. VSR-UAiC employs an end-to-end reinforcement learning (RL) algorithm to train agents for obtaining the optimal upload ABR strategy. A series of experiments have demonstrated the superiority of the VSR-UAiC framework over alternative solutions in terms of latency, cost, and system capacity.

Video content delivery networks (CDNs) have gained widespread adoption due to their capability to deliver low-latency and high-concurrency services. Through the routing of requests, the optimal media source can be discovered, thereby optimizing resource utilization and improving user experience. However, current request routing mechanisms neglect the differences among various media services, leading to routing decision schemes that cannot meet the diversified service demands. To bridge this gap, we introduce a Deep learning-based approach that leverages the Coordination between Service and Network for Differentiated Request routing in video CDN (DCSN-DR). First, we develop an Adapted Savitzky-Golay (SG) filter-Long Short-Term Memory (LSTM)-Attention (ASLA) prediction algorithm which combines noise removal, feature extraction, and outpusts weight allocation. Furthermore, we propose an innovative differentiated routing algorithm that capitalizes on the coordination between service and network. The proposed solution transcends the traditional emphasis on network-centric metrics by adopting a comprehensive perspective that encompasses both service-level and network-level factors. We evaluate the performance of the proposed DCSN-DR algorithm using a real-world traffic dataset obtained from an information service provider (ISP). The results demonstrate that DCSN-DR outperforms existing algorithms in terms of user traffic and throughput rate of total nodes.

Content Delivery Networks (CDNs) are widely used for their ability to provide highly concurrent services with low latency, and the large amount of log data generated by CDN operations helps information service providers (ISPs) optimize CDN resource allocation, evaluate performance, and analyze operational conditions. Unfortunately, CDN concurrency studies based on real runtime data have been neglected. In addition, current neural network-based time series prediction methods often degrade model prediction performance due to inter-task interference when applied to multi-task long-term data. Consequently, this paper addresses this gap by presenting a novel hybrid network concurrency prediction framework with Multi-Task Learning (MTL), Moving Average Extended Kalman (MAEK) filter, Long Short-Term Memory (LSTM) network, and the Attention mechanism, denoted as MMLA. In this framework, the MAEK filter is employed to eliminate noise from the original data. Subsequently, LSTM with MTL is harnessed to capture both long and short-term data dependencies while considering two key features: the day of the week and the type of service. Finally, the Attention mechanism assigns weights to critical time steps of various tasks, thus mitigating inter-task interference and enhancing prediction accuracy and robustness. Empirical assessments are conducted using a sizable dataset from a real ISP in a single province. The experimental results show that the proposed MMLA outperforms the existing solutions, in terms of root mean square logarithmic error (RMSLE) and mean absolute percentage error (MAPE).

With the proliferation of Internet of Things (IoT) devices and the exponential growth of data generated at the network edge, there is a pressing need for efficient task offloading strategies in edge-cloud collaborative systems. In this study, we address the optimization of task offloading decisions and computation resource allocation in a multi-user computation resource network comprising edge servers and a centralized cloud server interconnected. Our objective is to minimize both time delay and energy consumption. We formulate the problem as an optimization task aiming to minimize the integrated cost of latency and energy consumption while satisfying the delay and computation resource requirements, resulting in a non-convex, NP-hard problem. To tackle this challenge, we propose a deep reinforcement learning approach, specifically the Actor-Critic based Task Offloading Optimization Network (ACTOON). Extensive simulations are conducted to demonstrate the superiority of ACTOON over other baseline methods.

Recently, live streaming technology has been widely utilized in areas such as online gaming, e‐healthcare, and video conferencing. The increasing network and computational resources required for live streaming increase the cost of content providers and Internet Service Providers (ISPs), which may lead to increased latency or even unavailability of live streaming services. The current research primarily focuses on providing high‐quality services by assessing the resource status of network nodes individually. However, the role assignment within nodes and the interconnectivity among nodes are often overlooked. To fill this gap, we propose a hierarchical game theory‐based live video transmission framework to coordinate the heterogeneity of live tasks and nodes and to improve the resource utilization of nodes and the service satisfaction of users. Secondly, the service node roles are set as producers who are closer to the live streaming source and provide content, consumers who are closer to the end users and process data, and silent nodes who do not participate in the service process, and a non‐cooperative game‐based role competition algorithm is designed to improve the node resource utilization. Furthermore, a matching‐based optimal path algorithm for media services is designed to establish optimal matching associations among service nodes to optimize the service experience. Finally, extensive simulation experiments show that our approach performs better in terms of service latency and bandwidth.

In the past IPTV CDN servers test of China telecom, the testers only conducted stress tests on a single device. The test model only takes into account the factors based on the characteristics of user concurrent traffic, not the end-to-end model and the replacement of video popularity based on user behavior. Therefore, the test results do not accurately reflect the performance of CDN devices in the real environment. According to the current network IPTV CDN networking architecture and service requirements, this paper customized the IPTV CDN service performance index architecture, and designed the end-toend IPTV CDN test bed. At the same time, based on the in-depth analysis and research of the IPTV user access log of a province and the change rule of video heat, this paper designs a user access model with a long-time stability, proposes the intelligent replacement algorithm of content heat, and deploys it in the IPTV CDN test bed. The high-fidelity IPTV CDN test bed proposed in this paper has been officially launched by China Telecom, and has undertaken the evaluation and test of many commercial CDN equipment, as the technical basis of CDN equipment procurement of China Telecom.

As 5G services continue to rapidly develop, operators are experiencing an increasing proportion of network traffic usage devoted to streaming content. Streaming media is particularly suited for transmission in multicast networks. This paper presents a low-latency live streaming system that leverages Bit Index Explicit Replication (BIER) multicast. Furthermore, we complete end-to-end push-pull stream verification within a BIER multicast laboratory to demonstrate the effectiveness of our design. To serve as a case study for this system, we deployed it within an enterprise by connecting internet streaming media to the system. This deployment resulted in significant reduction of resource consumption during internal meetings and training sessions, which were previously using external bandwidth. Additionally, this approach allowed for unified management of streaming media gateways across multiple branches of the enterprise, even those scattered across different geographic regions, optimizing resource allocation. Thus, the system now supports daily operations of internal meetings and training sessions.

Live video streaming services have experienced significant growth, and it has imposed more stringent requirements on current media delivery networks. However, traditional media distribution methods have difficulty meeting the low latency and high bandwidth requirements of emerging live-streaming services. Therefore, for the live video service scenario, we propose a video computing and network convergence (VNCN) model containing multiple nodes and multiple users, which integrates the network and computing resource occupancy of each node in the system and the impact of the media delivery method on the live service. In addition, we also propose a resource-aware dynamic scheduling (RDS) algorithm, which dynamically schedules user requests based on the resources of each node to maximize user Quality of Service (QoS). Finally, the experimental results show that, compared with the commonly used Round-robin (RR) algorithm and the K-Nearest Neighbor (KNN) algorithm, our system can not only provide users with a high QoS live streaming service but also ensure a more balanced load on the resources of each node.

With the rapid development of virtual reality (VR) technology and the increasing demand for higher video quality, streaming videos imposes stringent requirements on the network. To solve these problems, the Field of View (FOV) streaming is one the popular solutions. In streaming aspect, it's common to provide video service over Internet Protocol Television (IPTV) platform. The application of 8K VR FOV in IPTV platform can reuse the existing Content Delivery Network (CDN) architecture and set-top boxes (STBs), thus significantly reducing the complexity of the system. Furthermore, it can also accelerate the popularization of VR technology, and improve the service development of IPTV platform. Currently, most of the VR FOV solutions in industry are based on Dynamic Adaptive Streaming over HTTP (DASH), which are difficult to apply in most IPTV platforms. Our Real Time Streaming Protocol (RTSP) 8K VR FOV video end-to-end solution for IPTV platforms significantly reduces bandwidth by about 60% compared to the traditional solution, and provides the same subjective quality. We evaluate and compare the performance of our solution with traditional 8K solution. The results show the superiority of the proposed architecture in terms of bandwidth, latency and CPU utilization.