The demand for immersive multimedia experiences has propelled the adoption of 4K 360-degree video streaming. However, delivering such high-resolution content imposes challenges related to bandwidth constraints and computational overhead. Adaptive streaming based on the user Field of View (FoV) offers a solution by providing high-quality video only in the viewed area and reducing the quality of peripheral regions, thereby optimizing bandwidth usage with low impact on user experience. This paper proposes a cloud-edge architecture for live encoding and streaming of 4K 360-degree videos using FoV-based adaptation within the Dynamic Adaptive Streaming over HTTP (DASH) framework. Leveraging cloud-native technologies like Kubernetes and Knative, the system employs a GPU-accelerated serverless architecture triggered by event-driven mechanisms. Experimental results demonstrate that the proposed system meets the latency requirements of live streaming, maintaining end-to-end processing times well below the 10-second threshold. Additionally, the adaptive streaming approach can achieve up to a 77% reduction in bandwidth consumption by adjusting the quality of video tiles based on the user’s FoV.

ABSTRACT Background Modern microservice architectures demand not only modularity, scalability, and maintainability but also adaptation to dynamic requirements. For applications deployed on Kubernetes, sidecar proxies have been used to separate the operational features (such as security, networking, or monitoring) from the application business logic by intercepting traffic to and from microservices or functions in serverless platforms. Existing proxy sidecar implementations such as Envoy offer these operational features as chains of logic that cannot be composed at deployment time. Objective This work introduces K8Sidecar, which leverages the operator pattern to dynamically integrate a chain of proxy sidecar containers into microservices or serverless architectures deployed on top of Kubernetes. Methods The study presents the architecture of the software and describes the Java and Go libraries provided to develop new proxy sidecars. Results We provide illustrative examples and evaluate the impact of the number of injected sidecars on cold‐start and latency and compare them with Envoy. Conclusion Results show that for up to 5 sidecars, the chain‐of‐sidecars approach (especially using the Go library) remains reasonably close to Envoy in terms of both cold start and latency.

In the digital era, video content dominates network traffic, with HTTP Adaptive Streaming commonly used by streaming services. Recently, novel video encoding schemes based on Deep Neural Networks (DNN) have been proposed. One of the proposals used features extracted from fixed length downscaled video segments. The DNN used as input the features and the desired Video Multimethod Assessment Fusion (VMAF), to estimate adaptively the Constant Rate Factor (CRF) to be applied to the original segment to attain the target VMAF. In this work we analyze the generalization capabilities of this trained DNN when instead of fixed length segments we consider scenes, that have variable duration. We show that the CRF is adapted per scene and per video and the final encoded videos have the requested VMAF. According to our tests, the encoded videos have an average absolute error of 1.34, being less than 2 and therefore not perceived by the end user. Besides, the size of the encoded videos decreases, up to 12% in the best case, compared to those obtained estimating the CRF from fixed length segments.

In the dynamic area of container orchestration, precise and timely metrics collection is essential for optimal resource management. PodInsights, an innovative Kubernetes metrics collector, surpasses existing tools by directly collecting pod CPU and memory usage data from the operating system at millisecond intervals. Deployed as a DaemonSet, PodInsights efficiently monitors selected pods based on configurable labels, offering unparalleled precision in capturing rapid fluctuations in resource consumption. Rigorous testing across different scenarios validates its accuracy, highlighting its potential to improve high-frequency data gathering in Kubernetes environments.

Abstract In recent years, Internet traffic patterns have been changing. Most of the traffic demand by end users is multimedia, in particular, video streaming accounts for over 53%. This demand has led to improved network infrastructures and computing architectures to meet the challenges of delivering these multimedia services while maintaining an adequate quality of experience. Focusing on the preparation and adequacy of multimedia content for broadcasting, Cloud and Edge Computing infrastructures have been and will be crucial to offer high and ultra-high definition multimedia content in live, real-time, or video-on-demand scenarios. For these reasons, this review paper presents a detailed study of research papers related to encoding and transcoding techniques in cloud computing environments. It begins by discussing the evolution of streaming and the importance of the encoding process, with a focus on the latest streaming methods and codecs. Then, it examines the role of cloud systems in multimedia environments and provides details on the cloud infrastructure for media scenarios. After doing a systematic literature review, we have been able to find 49 valid papers that meet the requirements specified in the research questions. Each paper has been analyzed and classified according to several criteria, besides to inspect their relevance. To conclude this review, we have identified and elaborated on several challenges and open research issues associated with the development of video codecs optimized for diverse factors within both cloud and edge architectures. Additionally, we have discussed emerging challenges in designing new cloud/edge architectures aimed at more efficient delivery of media traffic. This involves investigating ways to improve the overall performance, reliability, and resource utilization of architectures that support the transmission of multimedia content over both cloud and edge computing environments ensuring a good quality of experience for the final user.

Nowadays, video content accounts for a large percentage of network traffic. Most streaming services use HTTP Adaptive Streaming by splitting the video in non-overlapping segments and by encoding each video segment independently (possibly with multiple representations to allow adaptation to the varying network conditions). In this work we propose an encoding scheme based on a Deep Neural Network (DNN), to perform a per-title and per-segment adaptive estimation of the encoding parameter to achieve a target video quality of the encoded video. A dataset has been prepared using 1212 segments obtained from 158 videos. The segments have been encoded with 19 Constant Rate Factor (CRF) values using the VP9 encoder, generating a total of 23028 encoded segments, and for each encoded video segment, its Video Multi-Method Assessment Fusion (VMAF) quality has been computed. Besides, from a 240p downscaled version of the segments a feature vector has been obtained, and an analysis of the dependency of the features with the resolution has been carried out. With this dataset a DNN has been trained to estimate the CRF to be applied to each segment to achieve a target VMAF quality. Results show that the trained network is able to provide the CRF value to be applied to each video segment to achieve the desired quality with low computational overhead. To validate the proposal, the network has been used to predict the CRF to encode 1840 two-second segments, not used during the training process, using the VP9 codec at Full High Definition (FHD) resolution, and four target quality values. Results show that the system adapts the CRF to each segment and that the final videos have a mean deviation of 1.84% with respect to the requested VMAF value.

Abstract Nowadays, the majority of Internet traffic is multimedia content. Video streaming services are in high demand by end users and use HTTP Adaptive Streaming (HAS) as transmission protocol. HAS splits the video into non-overlapping chunks and each video chunk can be encoded independently using different representations. Therefore, these encode tasks can be parallelized and Cloud computing can be used for this. However, in the most extended solutions, the infrastructure must be configured and provisioned in advance. Recently, serverless platforms have made posible to deploy functions that can scale from zero to a configurable maximum. This work presents and analyses the behavior of event-driven serverless functions to encode video chunks and to compute, optionally, the quality of the encoded videos. These functions have been implemented using an adapted version of embedded Tomcat to deal with CloudEvents. We have deployed these event-driven serverless pipelines for video coding and quality metrics on an on-premises serverless platform based on Knative on one master node and eight worker nodes. We have tested the scalability and resource consumption of the proposed solution using two video codecs: x264 and AV1, varying the maximum number of replicas and the resources allocated to them (fat and slim function replicas). We have encoded different 4K videos to generate multiple representations per function call and we show how it is possible to create pipelines of serverless media functions. The results of the different tests carried out show the good performance of the serverless functions proposed. The system scales the replicas and distributes the jobs evenly across all the replicas. The overall encoding time is reduced by 18% using slim replicas but fat replicas are more adequate in live video streaming as the encoding time per chunk is reduced. Finally, the results of the pipeline test show an appropriate distribution and chaining among the available replicas of each function type.

Video content represents a high percentage of the traffic on Internet with an increasing number of platforms offering live and on demand video. Most platforms use HTTP Adaptive Streaming (HAS) to deliver their content. In HAS, the provider splits the videos in segments and each segment is offered in multiple representations. The player can dynamically request the appropriate representation for the next segments depending on the varying conditions. This alleviates problems such as initial delay or stalling, thus providing a better quality of experience to users. These segments can have a fixed duration or can be adapted to match scenes up to a maximum duration of typically 10 seconds. In this work, we study the effects in time, quality and size of the encoded videos when downscaling the video to obtain the scenes. The experiments are performed using two codecs: H.264 and VP9, and using 10 videos with 4K resolution with a duration of more than 150 seconds. The videos are encoded using fixed segments and variable segments based on scenes (obtained from the downscaled video). Results show that the use of downscaling to obtain the scenes has a small impact on the final quality, while reduces the total time, the consumption of computational resources, and the size of the encoded video.

Today, more than 85% of Internet traffic has a multimedia component. Video streaming occupies a large part of this percentage mainly because this type of content is provided by the most used applications on the Internet (e.g. Twitch, TikTok, Disney+, YouTube, Netflix, etc.). Most of these platforms use HTTP Adaptive Streaming (HAS) to send this media content to end users in order to ensure a good quality of experience (QoE). But, this QoE should be guaranteed from the video to be transmitted, i.e., the video should have an adequate quality by minimizing the bitrate before transmission. In order to solve this issue, we present a system capable of encoding a video in several resolutions given the desired value of an objective metric. Our system includes the objective metric in the encoding loop in order to maintain the quality in all segments. This system has been tested with three video and five resolutions for each video. Our proposal provides improvements of more than 10% in terms of video size and with similar coding times when compared with a fixed Constant Rate Factor (CRF) encoding. A visual comparison between our proposal and a fixed CRF encoding can be seen at: https://links.uv.es/jgutierr/multiresQ

The emerging Fifth Generation (5G) mobile networks have been attracting enormous attention from various stakeholders around the world. In particular, in the research community, prototyping 5G infrastructures and deploying 5G services have gained gears recently towards realising market-oriented 5G trials. However, accessing to and programming on real-world 5G infrastructure is almost prohibitive for most 5G researchers especially in academia. Therefore, it is critical to build realistic yet cost-efficient 5G infrastructure emulators for 5G research labs to enable credible 5G research activities. This paper proposes such a 5G infrastructure emulator that is able to emulate a realistic 5G network in a lab setting based on a small number of commercial-off-the-shelf servers by leveraging virtualization and other technologies. Moreover, this emulator allows a service provider to automatically deploy 5G services from `empty' machines through advanced automation. The emulation platform is described in details with the 5G infrastructure and service deployment procedure highlighted. Empirical results are presented to show the performance of the proposed emulator.

The next generation mobile networks 5G are currently being intensively developed and standardized globally, with commercial prototyping 5G connections already emerging. At the 5G system level, one of the Key Performance Indicators (KPIs) defining 5G is substantially reduced service creation time for 5G network operators and/or service providers to increase the system efficiency and thus reduce operational costs. In this work, we focus on realize this challenging KPI in terms of speedy creation of monitoring services for 5G operators from scratch (no operating system pre-installed). A new 5G infrastructure orchestrator UWSIO is proposed to achieve fully automated deployment of 5G monitoring services. The architecture of this orchestrator is presented, which is compliant with the ETSI MANO standard to deal with both physical and virtual resources towards establishing the services running over the infrastructure. The proposed orchestrator is implemented in a real-world testbed and the implementation details are provided. Experimental results demonstrate that the performance of the design and implementation of this orchestrator is able to meet the KPI requirement for 5G operators.

With the advances of new‐generation wireless and mobile communication systems such as the fifth‐generation (5G) mobile networks and Internet of Things (IoT) networks, demanding applications such as Ultra‐High‐Definition video applications is becoming ever popular. These applications require real‐time monitoring and processing to meet the mission‐critical quality of service requirements and are expected to be supported by the emerging fog or edge computing paradigms. This paper presentsNFVMon, a novel monitoring architecture to enable flow monitoring capabilities of network traffic in a 5G multioperator mobile edge computing environment. The proposedNFVMonis integrated with the management plane of the Cloud Computing.NFVMonhas been prototyped and a reference implementation is presented. It provides novel capabilities to provide disaggregated metrics related to the different 5G mobile operators sharing infrastructures and also about the different 5G subscribers of each of such mobile operators. Extensive experiments for evaluating the performance of the system have been conducted on a mid‐sized infrastructure testbed.

The progress in realizing the Fifth Generation (5G) mobile networks has been accelerated recently towards deploying 5G prototypes with increasing scale. One of the Key Performance Indicators (KPIs) in 5G deployments is the service deployment time, which should be substantially reduced from the current 90 hours to the target 90 minutes on average as defined by the 5G Public‐Private Partnership (5G‐PPP). To achieve this challenging KPI, highly automated and coordinated operations are required for the 5G network management. This paper addresses this challenge by designing and prototyping a novel 5G service deployment orchestration architecture that is capable of automating and coordinating a series of complicated operations across physical infrastructure, virtual infrastructure, and service layers over a distributed mobile edge computing paradigm, in an integrated manner. Empirical results demonstrate the superior performance achieved, which meets the 5G‐PPP KPI even in the most challenging scenario where 5G services are installed from bare metal.

Monitoring of cloud computing infrastructures is an imperative necessity for cloud providers and administrators to analyze, optimize and discover what is happening in their own infrastructures. Current monitoring solutions do not fit well for this purpose mainly due to the incredible set of new requirements imposed by the particular requirements associated to cloud infrastructures. This paper describes in detail the main reasons why current monitoring solutions do not work well. Also, it provides an innovative monitoring architecture that enables the monitoring of the physical and virtual machines available within a cloud infrastructure in a non-invasive and transparent way making it suitable not only for private cloud computing but also for public cloud computing infrastructures. This architecture has been validated by means of a prototype integrating an existing enterprise-class monitoring solution, Nagios, with the control and data planes of OpenStack, a well-known stack for cloud infrastructures. As a result, our new monitoring architecture is able to extend the exiting Nagios functionalities to fit in the monitoring of cloud infrastructures. The proposed architecture has been designed, implemented and released as open source to the scientific community. The proposal has also been empirically validated in a production-level cloud computing infrastructure running a test bed with up to 128 VMs where overhead and responsiveness has been carefully analyzed.