Dynamic service function chain placement in mobile computing: An asynchronous advantage actor-critic based approach

Abstract

Internet of Things (IoT) devices are constantly sending data to the cloud. The resource-rich cloud computing paradigm provides users with significant potential to reduce costs and improve quality of service (QoS). However, the centralized architecture of cloud data centers and thousands of miles away from clients has reduced the efficiency of this paradigm in delay-sensitive and real-time applications. In order to get over these restrictions, fog computing was integrated into cloud computing as a new paradigm. Without using the cloud, fog computing can supply the resources needed for IoT devices at the network's edge. Delay is thereby decreased because processing, analysis, and storage are located closer to the clients and the areas where the data is created. In Mobile Edge Computing (MEC) networks, this study sets up an architecture based on Deep Reinforcement Learning (DRL) to deliver online services to end users. We introduce a DRL-based method named DPPR for Dynamic service function chain (SFC) Placement that uses Parallelized virtual network functions (VNFs) and seeks to optimize the long-term expected cumulative Reward. Online service provider DPPR can accomplish processing acceleration through parallel VNF sharing. In addition, by extracting the distribution of initialized VNFs, DPPR improves the capacity to handle subsequent requests. The conducted simulations demonstrate the efficacy of the proposed method, so that the average number of accepted requests is improved by about 11.7%.