The challenge of end-to-end encryption of YouTube and some other services

[mouakter13]

While edge computing is exciting for its ability to bring content to remote internet users, it is also well-suited to collecting data and information in these communities, especially from sensor networks. Edge servers on cell towers empowered with theoretically endless memory and computational power are very well-placed to collect and analyze data from locally deployed sensors. On the one hand, users’ phones are great data collectors: Modern smartphones contain approximately 20 different sensors. For example, their cameras can collect data on sunlight and their accelerometers can observe movements of people in a community. Yet there are far more uses for sensors, particularly those designed specifically for agriculture, the main form of employment in many developing countries. These may include monitors to detect water levels, the proportion of time when the sky is sunny or cloudy, soil acidity levels, livestock health and more. Having enough of this data stored on a local edge server could allow artificial intelligence running on the edge server to suggest when farmers should plant, whether the fertilizer they used was working and so on. Better data on crops could also help inform customized financial products timed to the harvest season. Farmers uploading data to an edge server could even agree to combine their data and re-sell it to researchers, or to the developers of relevant machine learning algorithms who are in search of training data. Development organizations are well-placed to serve as partners in these efforts, working with both local communities and companies looking to test their business models in under-resourced environments.

The growth of low earth orbiting satellites

The services and price points of low earth orbiting (LEO) satellites promise to be complementary to edge computing. These satellites aim to offer service in remote locations, and their reach is rapidly expanding: March saw the 41st launch of Starlink satellites, bringing the total deployed to over 2,000 out of a potential 42,000. While still pricey, the service offers a median download speed of 104 megabits per second, as measured by Ookla, the operator of an internet speed test. OneWeb, which had gone bankrupt in March 2020 but got back in the running thanks to investments by the U.K. government and Bharti, trails in terms of numbers but not ambition. And Amazon has two complementary programs, the nascent Project Kuiper LEO satellite network and AWS Ground Station, which offer (quite literal) vertical integration by delivering content via their own satellites to their own ground station, with onward distribution per customer demand. It is easy to imagine a combination of these two services working with edge networks in remote locations. In light of this momentum, in September 2021, two different satellite groups began calling for an open standard for satellite equipment, allowing for interoperability between platforms. Their ideas echo the thinking behind the creation of the Android operating system, as well as the systems created under the Telecom france whatsapp number data Infra Project, such as OpenRAN. Open standards should help reduce the cost of connecting edge systems to satellite backhaul.


While content caching sounds simple and cost-effective, there is one major problem for much of the content that people would likely want to cache—including content from sources such as YouTube: This content is encrypted end-to-end to protect intellectual property. This means cache servers cannot know what the content is, cannot identify which content multiple people are accessing, and cannot cache it. N50 member PicoNETS is trying to tackle this problem by linking up with the AWS Wavelength service, which “embeds AWS compute and storage services within 5G networks, providing mobile edge computing infrastructure for developing, deploying and scaling ultra-low-latency applications.” Since AWS in this case would be both a network AND a content provider, this linkage would make computing power at the edge much more powerful and accessible: A user would have just one company managing the entire content value chain, so if encryption is used, that company could cache its own content. In October Google provided further evidence of this trend, when it unveiled its competing Distributed Cloud service, which provides similar services to MNOs and other internet service providers. This may suggest a need for consumers to have access to multiple vertically integrated content providers (similar to the relationships many of us have with multiple streaming providers, like Netflix, Disney+, etc.). However, it may be the case that content providers will ultimately opt to collaborate to save on content delivery costs. Even if these companies’ initial focus is primarily 5G, the technology would likely be configurable for lower bandwidth environments. And try to imagine Amazon combining its Base Station and Wavelength businesses: They control so much content, these advancements that allow them to deliver their own content might sideline traditional carriers.