Research On Deployment Of Multi-Instance Service Chain For Edge Environment

In recent years,Internet services have increasingly changed people's lifestyles.Edge computing,as a new technology that responds to the increasing demand for realtime processing by users,provides enterprises with low latency for online services.Unfortunately,overloading any edge device in the edge environment will result in reduced service performance and even unavailability of the service,causing huge losses to enterprises.Therefore,for enterprises,the efficient deployment of edge services becomes very important.With the continuous development of network function virtualization technology,enterprises can deploy network functions on virtual instances,which decouples network functions from physical hardware devices and is a prerequisite for rapid development of edge computing.The rapid development of 5G technology further shows that edge computing and network function virtualization technologies have great potential to provide low-cost and low-latency services.As a carrier carrying virtual network functions,edge devices usually have limited resources.In order to cope with the possible shortage of equipment resources,enterprises must make a more fine-grained measurement of the resource consumption of network functions in the service chain.Therefore,some studies have found that through the benchmark experiment,there will be communication overhead that cannot be ignored between service chain network functions.On the other hand,in order to improve the stability of services and reduce the load of edge devices,enterprises usually deploy the same network function in virtual instances on different edge devices to provide users with highly available and low-load services.Based on the above facts,in order to further reduce the load on the edge devices,this paper deploys any network function of the service chain in virtual instance on different edge devices based on the fine-grained service chain load model,and then proposes a multi-instance service chain deployment problem under delay constraint.The first work of this paper proposes an online deployment algorithm for multiinstance service chain.It has the following contributions.First,we conduct an abstract modeling of the edge environment and multi-instance service chain deployment,and propose the edge environment model and the virtual instance communication diagram,as well as the edge device load model and the delay constraint model.Second,we design and implement an algorithm framework(DS-PC-ND)for solving the online deployment problem of multi-instance service chain,this framework includes three components: delay path search algorithm based on pruning search strategy,deployment path selection algorithm based on nested Top K strategy,and network function deployment algorithm based on greedy strategy.Third,experimental results show that the algorithm DS-PC-ND can better than the benchmark algorithm that ignores the communication overhead between service chain network functions by 10% of the CPU load rate of the edge device.The second work of this paper solves the optimal deployment result of the multiinstance service chain and optimizes the solution efficiency.It has the following contributions.First,we use nonlinear constraints to transform into linear constraints and other optimization strategies to convert mixed integer quadratic programming problems to mixed integer linear programming problems,this makes it possible to quickly solve the optimal deployment results in a longer time dimension.Second,experimental results show that the deployment result of the online algorithm DS-PC-ND is close to the optimal deployment result.