Title: EdgeTube – Smart Edge Computing-based video acquisition for efficient distribution Scalable edge computing based smart video acquisition for efficient distribution
Company: Ubiwhere, Lda
When (time-plan): November 2018 to May 2019
EdgeTube, from Ubiwhere, is an H2020 FLAME 1st Open Call winner hoping to create a new business case for efficient and low-cost crowdsourced media acquisition and distribution, using key technology such as Edge Computing. The designed experiment is targeting both Mobile Network Operators, Service Providers and even Event Organizers. With this solution, and in the scope of the company’s Smart Lamppost solution (http://smartlamppost.com), such actors can not only use this smart urban furniture to extend the network coverage and capacity to serve the the event’s attendees (with built-in Small Cells), but also to provide value-added solutions, by deploying and re-using Edge Computing hardware to pre-process and pre-filter media-related content. Being able to do so allows such actors to (i) save costs on bandwidth (by avoiding unnecessary traffic in the backhaul) and computational processing power and (ii) to allow for a better QoE experiment for cloud-connected consumers of such media streams. In essence, and focusing on the end-user, the solution allows for viewers to watch different segments of the event (live coverage of specific areas or events), providing that the system incentivizes local attendees to live stream different content with the best possible quality. Due to the system’s intelligence, the solution hopes to provide viewers with a global coverage of the event, while ensuring the best possible quality for each segment, with very little interruptions due to the built-in automatic failover mechanism (pool of backup streams).
Scope/objective of the trial
To demonstrate the system’s capabilities, a real trial has been set-up at Bristol’s premises, using the dedicated infrastructure (WiFi and edge computing). For this, different services (as depicted in the figure below) have been deployed at the Edge and two different experimenters have used our EdgeTube mobile app to stream the acquired content, thus simulating real attendees of a live concert, for instance.
In simple terms, whenever a user starts sharing video content, the system periodically checks for this content’s quality and user’s orientation: if other users are streaming following the same orientation, the system compares the quality of each one, thus deciding which one should be available for consumption; if it is unique, they are immediately available for consumption. Whenever the system detects there is another higher quality stream, the user is advised of such and, should the user not explicitly stop the streaming, the system periodically checks for newer chunks of video, thus being constantly monitoring for better streams.The diagram below briefly summarised this process.
While in Bristol, the system proved its capabilities. Even though specific bugs have been detected, it performed as intended. In essence, EdgeTube allowed for:
- Saved bandwidth: leveraging Edge Computing, the system pre-processes the media quality before re-transmitting it all the way to a cloud-environment (available for further mass consumption). Instead of re-transmitting all of the acquired media, only the best stream for a given orientation is made available.
- The services scaled accordingly, using FLAME’s scalability mechanisms based on specific parameters such as CPU usage
- Media analysis at the edge took less than 5 seconds, offer a good user experience (subjective, but based on qualitative feedback analysis)
To know more about Be Memories experiment, read also their blog post.