Energy-metering and its reduction in a video CDN over post-5G networks

1. Context
The rapid scaling of IT infrastructure is raising a persisting concern about the needed energy to power them and their actual carbon footprint impact. Besides, powering IT infrastructure constitutes a signification portion of the operational budget that operators seek to reduce by all means. This is particularly true for video CDN providers that are dealing with an increasingly large volume of media content and streams forcing them to scale up their processing, storage and network capacities in order to satisfy that demand. This trend is particularly accelerated due to the blooming of remote Extented Reality (XR) services which are resource-intensive. These facts inevitably imply a surge in the consumed energy by the delivery infrastructure. Given that situation, video CDN providers are actually attentive to energy-saving mechanisms and procedures minimizing their operational power requirements. Moreover, it is imperative to develop new energy-aware video CDN solutions that would be considered and integrated in the scope of post-5G networks.

2. Objectives and challenges
In earlier network generations, video CDN has been constructed as a resilient overlay network that is built upon an existing network infrastructure. It is known that overlay solutions requiring coherent global state are generally energy-draining. There are two main determinants of energy consumption: (a) the number of peers and content size and (b) network and content dynamics. Both of them are expected to increase dramatically in the near future. The energetic aspect has been largely overlooked in the literature as most
attention is paid to improve performance metrics such as speed, quality and reliability of provided video content. That is why, we aim in this work to perform a thorough energy usage audit of the actual video CDN. Based on that, we are going to develop a new and specific energy-saving video CDN solution that could be easily integrated in post-5G networks. To do that, we have identified the following two challenges that will be addressed :

Challenge 1: Fine-grained metering, analysis and modeling of energy consumption of video CDN
The end-to-end energy usage metering of video CDN is a real challenge [1]. This is particularly due to the geographical distribution of video peers that are typically placedwithin autonomous ISP systems, where video providers have a limited control. Moreover, video streams exchanged over a video CDN are not isolated from the remaining streams, which make quantification of energy consumption caused by CDN traffic very subtle.

Challenge 2: Energy-aware content synchronization policy across software-based video peers
The content synchronization is a key aspect of video CDN that can be performed following different policies, such as cooperative push-based and cooperative or non-cooperative pull-based. They have been designed in order to maintain a certain level of QoE (Quality of Experience) and to provide an optimal resource usage. The challenge here is to introduce energy-saving mechanisms that keep actual performance while reducing consumed energy. These mechanisms include, among other, IA-based garbage collector, succinct content versioning systems, and sophisticated peers management.

3 . Action plan

• Perform a state of the art of existing energy-metering approaches and techniques of cloud-based and overlay networks [2,3]. We plan to compare different solutions in term of accuracy, overhead and parsimoniousness. Next, we will examine and discuss their application for video CDN.
• Propose a dedicated energy-metering approach and models for video CDN. To do that, we plan to propose an optimized and fine-grained metering strategy under different post-5G scenarios that will be evaluated using NS3.
• Propose new in-band and/or out-of-band synchronization protocols that optimize energy usage while keeping a certain QoE level and guarantees a global coherence across video peers. We will consider both centralized and distributed energy control solutions. We will consider software-based peering server that provides a broad range of administration actions.