Abstract
This publication describes systems and techniques for adaptive remote execution of computing tasks from a mobile computing device, such as a smartphone. The computing resources required by applications (and/or computation blocks thereof) can vary significantly due to workload and other factors. Execution of resource-intensive computation blocks can quickly drain battery power and overwhelm limited on-device resources, resulting in poor user experience. In some circumstances, offloading compute tasks through low-latency, high-speed networks can be advantageous. The described systems and techniques determine optimal offload decisions for respective compute blocks; the decisions may be adapted to current conditions (e.g., network connectivity, network reliability, device load, and/or the like), characteristics of the respective compute blocks (e.g., battery consumption, execution time, and/or the like), and so on. Offload decisions may be determined in accordance with a machine-learned modeling approach based on an artificial neural network, Contextual Bandits, Reinforcement Learning, and/or the like.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Carbune, Victor and Damian, Alexandru, "Adaptive Remote Execution", Technical Disclosure Commons, (March 16, 2020)
https://www.tdcommons.org/dpubs_series/3023