Key points are not available for this paper at this time.
Edge computing is a promising technology, especially for offloading users’ computationally heavy tasks. The close proximity of edge computing units to users minimizes network latency, thereby enabling delay-sensitive applications. Although optimal resource provisioning and task offloading in edge computing are widely studied in the literature, there are still some critical research gaps. In this study, we propose a traffic-aware optimal association and task-offloading approach. The proposed method does not rely solely on the average rate of offloading requests, which can differ from actual values in real time. Instead, it uses an intelligent, high-precision prediction model to forecast future offloading requests, allowing resource provisioning to be based on future sequences of requests rather than average values. Additionally, we propose an optimization-based approach that can meet task deadlines, which is crucial for mission-critical applications. Finally, the proposed approach distributes the computing load over multiple time steps, ensuring future resource scheduling and task-offloading decisions can be made with a certain level of flexibility. The proposed approach is extensively evaluated under various scenarios and configurations to validate its effectiveness. As a result, the proposed deep learning model has resulted in a request prediction error of 0.0338 (RMSE). In addition, compared to the greedy approach, the proposed approach has reduced the use of local and cloud computing from 0.02 and 18.26 to 0.00 and 0.62, respectively, while increasing edge computing usage from 1.31 to 16.98, which can effectively prolong the lifetime of user devices and reduce network latency.
Nugroho et al. (Wed,) studied this question.