Service chain parallel deployment method and device based on deep reinforcement learning

Abstract

The invention discloses a service chain parallel deployment method and device based on deep reinforcement learning, and the method comprises the steps: carrying out the mathematical modeling of an offline service chain deployment problem, so as to obtain a mathematical formula of the service chain deployment problem; selecting a placement server position for the shared VNF in all the service chains according to the mathematical formula, and selecting the server position of the shareable VNF through the DQN in deep reinforcement learning to generate a plurality of sub-service chains; and connecting the plurality of sub-service chains into a complete service chain through a shortest path principle, and selecting to deploy a server for the VNF without the appointed placement position. According to the method, the problem that the VNF in the service chain is ignored due to serial deployment and the unreasonable distribution is caused by the correlation between the service chains is solved,the sharing rate and the utilization rate of resources are effectively improved, deep reinforcement learning is adopted, and the calculation complexity is reduced.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a method and device for parallel deployment of service chains based on deep reinforcement learning. Background technique [0002] In today's enterprise and data center networks, end-to-end network service deployment usually requires a variety of network functions, including firewalls, load balancers, and deep packet inspection, etc., and service traffic needs to pass through a series of Network functions, these ordered network functions form a service function chain (Service Function Chain, SFC). The emerging network function virtualization (Network Function Virtualization, NFV) technology changes the implementation of these network functions by migrating them from dedicated hardware to commodity servers, that is, software-based proprietary hardware, and in NFV It is called a Virtual Network Function (VNF). The development trend of NFV enables operators to more fl...

AI Technical Summary

Problems solved by technology

However, due to too much slack, the computing power of the server in the above solution may exceed the actual physical resource capacity by up to 16 times

[0012] (2) Limited to heuristic algorithms, although server resources and link bandwidth resources are jointly considered, most of these methods are limited to heuristic algorithms

The heuristic method converges faster, but iteratively solves the problem, which affects the quality of the solution and increases the time of solution

[0015] (3) Serial placement lacks consideration of the relevance of the service chain

Even in the solution of the service chain placed in serial order, there has been a solution design for sharing with the VNFs in the previously placed service chain, but the sharing based on serial placement is very limited, because the sharing of serial placement can only Arrange according to the layout of the service chains that have been placed before, without taking into account the service chains behind

Images