Research On Joint Allocation Mechanism Of Computing Resource And Wireless Resource Based On Satellite Slicing Network

With the explosive growth of data traffic,satellite network,as one of the most potential communication networks,have problems such as single function,insufficient flexibility,and long transmission delay.At the same time,in order to make more effective use of increasingly scarce satellite resources,for the current resource allocation algorithms’ issues,such as consider service priorities and channel conditions insufficiently,waste of resources,and inability to achieve efficient utilization of one resource type.We need to put forward own ideas about satellite network architecture and resource allocation plan.Through in-depth research on bandwidth and computing resource’s allocation,this article analyzes how to solve problems related to resource allocation and improve system performance based on adapting to the scenario.First,based on the research on the existing satellite network architectures and key technologies in 5G,this paper designs a new architecture of software-defined satellite networks based on network slicing and mobile edge computing.This architecture separates data plane and control plane by software-defined network technology,then realizes network and hardware update independently;using network slicing technology to achieve faster and more flexible differentiated service provision;in delay-sensitive slices,deploying mobile edge computing servers to reduce network traffic pressure and delay.Secondly,based on the new network architecture proposed in this paper,a dynamic bandwidth slicing strategy based on reinforcement learning is proposed.Different from other sliceing strategies,this strategy uses the reinforcement learning algorithm to learn the knowledge of the wireless environment dynamically,so as to update the number of resources allocated to slices dynamically;by establishing a utility model that comprehensively considers users’ fairness,QoS,and channels’ quality,we make this model as RL’s reward function to achieve the goal of maximizing system utility;at the same time,the strategy uses an access control mechanism in the utility function and Q-learning’s action to avoid invalid allocation.Simulation proves that the proposed dynamic bandwidth slicing strategy has a good effect in improving resource utilization and system utility.Finally,in the delay-sensitive slices of the new network architecture proposed in this paper,a joint allocation strategy of computing resources and bandwidth resources based on deep reinforcement learning is proposed.As the scenario served by this strategy is more complicated,this strategy uses deep reinforcement learning for two type of resources’ dynamic allocation.In order to achieve a fine-grained combination of computing resources and bandwidth resources,this strategy studies the impact of these resources on the system’s utility and on each other’s utility.a utility model of joint allocation of two resources is established.At the same time,based on the access control idea in the dynamic bandwidth slicing strategy,this strategy improves to avoid invalid resource allocation and select computing tasks to be transmitted to the center,it’s one of the reason that why system’s utility and resource utilization are improved.Simulations also prove the effectiveness of this strategy.