Distributed Computing Using Dynamic Local Network of Edge Devices

Conventional cloud computing operation involves direct communication between nodes in the cloud and individual client devices. Such an approach is inefficient for handling situations in which cloud nodes or client devices may be occasionally offline. This disclosure describes an architecture that provides operational flexibility by forming a dynamic local network of multiple client devices associated with a user that provides redundancy and avoids computing bottlenecks. The dynamic local network is used to distribute computations in a manner that can overcome the bottlenecks that result from a cloud node or client device being offline. When a cloud node is offline, one or more client devices can hold and run the required computation. Similarly, if a targeted client device is offline, a cloud node can allocate the task to another client device associated with the user. Such operations can lower feature subscription costs and reduce latency, thus enhancing the user experience