Multi-Service And Multi-Tier Edge-Cloud Continuum Planning

This paper presents optimal planning of multiservice and multi-tier edge-cloud environments. The goal is to evaluate the whole deployment of compute continuum, i.e. the type and number of processing devices and their pairing with specific tier and task based on processing and latency requirements. Especially, this pairing of each task with specific processing device and processing tier is extracted under cost and power minimization objectives. The proposed planning approach is generic therefore applicable to any type of services. In the current work it is applied and validated in the AI inference service tasks. Results show the trade-off among the performance and execution time, i.e. the proposed optimal mechanism outperforms a heuristic scheme (providing a feasible solution of the studied problem), while the latter is less time consuming. Finally, further analysis of trade-offs subject to the execution of these two approaches is presented considering the needs of system’s service provider and users.