ABSTRACT
Abstract
This work jointly considers dynamic function configuration, function-download delay, multi-slot task offloading, and energy harvesting in serverless edge computing. It formulates a MILP that maximizes the number of processed tasks by coordinating offloading, function configuration and scheduling, download decisions, energy use, and channel allocation. A distributed D-TOEH protocol is then proposed for scalable operation and is reported to achieve 71.36% of the MILP benchmark on average with low polynomial complexity.