Research On Service Function Chain Mapping Strategy

At present,the data center network deploys a large number of servers and network devices to provide various network service functions(firewalls,DPI,IDS).Most traffic in a data center network is handled by multiple server functions to form a traditional network service chain.However,traditional network equipment have such defects as strong coupling between devices,poor flexible deployment,and troublesome expansion and upgrades.With the advent of NFV technology,it is possible to virtualize network devices on traditional commodity servers without using expensive dedicated equipment.Software-defined networking(SDN)supports simpler network management and setup it can separates network control functions from physical networks.Based on this,this thesis proposes two SDN/NFV-based service function-chain mapping strategies,which are used for service function allocation and deployment mapping and service link mapping.1.Due to the large amount of redundancy of data center network servers and the large cost of mapping SFC service functions,this thesis propose a service functionchain mapping algorithm(Splitting/Aaggregating based service function chain mapping,SA-SFCM).First of all,for this problem,we construct a corresponding system model,and then build the mapping problem into a 0-1 plan based on the relationship between balanced energy consumption and service performance.In addition,We proposed a "Relevance" indicator to measure the traffic intensity between service function,analyzes the characteristics of the nodes to be merged,and uses this to design a consolidation strategy that minimizes the traffic between servers.Then,the merged function body is split and mapped to improve the utilization of fragment resources and reduce the amount of server redundancy.The mutual traffic relationship between the split mapping objects is analyzed,and a suitable splitting strategy is proposed to reduce the mutual traffic.The final simulation results show that the SA-SFCM algorithm is superior to some existing algorithms in energy consumption and delay.2.For the route mapping problem of the service function chain in the data center,this thesis propose a Q-learning algorithm based on enhanced learning to solve the problem.First of all,for this algorithm,we establish a corresponding system model,give the SFP(service function path)and RSP(rendered service path)model.Then the node load,link load and delay of the route map are constituted a minimum weight optimization problem.We use the Q-learning algorithm to perform simultaneous SFP The standard chosen by the RSP is to map the links with the lowest link utilization.The final simulation results show that the Q-learning-based routing algorithm performs well in terms of load balancing and delay.