Ph.D. Dissertation:

An Autonomous Planning Model For Deploying IoT Services In Cloud-Fog Computing

Due to the ever-increasing use of Internet of Things and their related data transfer in a centralized cloud, we face more delay, bandwidth and cost. Although the cloud computing model has significant potential to reduce costs, the centralized architecture of cloud computing reduces the efficiency of cloud computing in delay-sensitive applications. Therefore, fog computing is a new paradigm in cloud computing to overcome these limitations. Cloud computing is able to provide the resources needed by receivers at the edges of the network, so data processing, analysis and storage are closer to the receivers' services and data creation locations, which reduces the cost of delay. In this thesis, using the competitive imperialist ICA algorithm as a meta-heuristic method and the A3C algorithm as a DRL learning method, solutions are presented to improve service deployment issues in fog computing. In the ICA method, the algorithm reduces the delay by prioritizing Internet of Things services and solves the service deployment problem as a multi-objective problem. In the A3C method, according to the objectives of minimizing the cost of delay under the included deadlines and resource limitations, it focuses on the deployment of resources. The results obtained from the simulation of the methods in an artificial fog environment show the improvement of the efficiency of the proposed algorithms compared to the latest algorithms and recent researches.

Keywords: IoT, Fog Computing, Service Placement, ICA, A3C