Network service function chain deployment method

Abstract

The invention discloses a network service function chain deployment method. The method comprises the the steps of 1) reading a service function chain request; 2) calculating a global resource capacityvalue of each node of the underlying network; 3) establishing a graph convolutional neural network; 4) selecting a candidate deployment set B={B1, B2,..., BN}; and 5) deploying the physical network according to the candidate deployment subset Bk with the maximum value function, and updating the underlying network information. According to the method, a node sorting algorithm is used for probabilistically generating a selected deployment set to reduce the time complexity, a time sequence difference algorithm is used for updating a candidate value function, the graph convolution neural networkis used for dynamically capturing network topology and residual resources, and an online deployment decision for real-time arrival is carried out on a network function service request according to anoptimal value function.

Description

technical field [0001] The invention relates to the field of network service function chains, in particular to a network service function chain deployment method. Background technique [0002] In traditional networks, service providers rely on middleboxes for network functions such as network address translation, intrusion prevention systems, firewalls, and load balancers. However, these network functions are directly connected into dedicated hardware, resulting in insufficient scalability, inflexibility, and high management costs in service deployment. To address these issues, Network Function Virtualization (NFV) transforms network functions from dedicated hardware to software middleboxes. NFV technology deploys network services or applications in the form of virtual network functions (Virtual Network Functions, VNFs), enabling flexible and scalable deployment and management. In order to compose complex services, network traffic usually needs to pass through a set of VNF...

AI Technical Summary

Problems solved by technology

[0003] The existing method has a deployment algorithm based on mathematics. The simulation results show that this algorithm can solve the offline deployment of SFC, but it needs to consume a lot of computing power. In addition, in real scenarios, the arrival time of SFC and the requested resources are unknown, so SFC deployment should be an online deployment problem. There are also deployment methods based on heuristics and metaheuristics. Although these methods can solve the above problems, they may fall into local optimal solutions. There are also some researches on deployment based on reinforcement learning. However, these methods either have a large search space, are not suitable for real-time deployment decisions, or rely on manual selection of feature learning algorithms to affect their effectiveness.

Images