Research On Online Deployment Algorithms Of Service-Function-Chains In Multi-Access Edge Computing

With the rapid development of mobile communication,artificial intelligence and chips,some highly real-time and complex applications,such as augmented reality,selfdriving,remote diagnosis,have gradually enter daily life.However,due to the limited resources of mobile terminals,it is impossible to process the large amount of data generated by mobile applications in a short period of time.Therefore,Multi-Access Edge Computing(MEC)and Network Function Virtualization(NFV)have been proposed.With the combining of MEC and NFV,how to deploy Service Function Chain(SFC)in the edge computing environment with limited computation and communication resources has become a new research hotspot.Here,two online algorithms are proposed for the Online SFC Deployment Revenue Maximization(OSDRM)problem to improve the service provider’s revenue while satisfying the task requirements,respectively.The specific works are as follow.1.An offline model of the OSDRM problem is constructed,and the OSDRM problem is proved as an NP-hard problem.Then,the task deployment cost model is formulated to calculate the deployment cost from the remaining computation resources of the edge server and the remaining bandwidth capacity of the link.After that,the OSDRM problem is converted into the Online SFC Deployment Profit Maximization(OSDPM)problem.Finally,the Online Profit Maximization Approximation(OPMA)algorithm is designed to solve the OSDPM problem,and its competitive ratio is proven.2.An offline model of LOSDPM is built for the Latency-aware Online SFC Deployment Profit Maximization(LOSDPM)problem by researching the latency constrains of tasks in MEC.Then,the Online Latency-aware Binary Search(OLBS)algorithm is designed to solve the LOSDPM problem.3.The proposed OPMA algorithm is compared with the online algorithm GreedyC and the offline algorithm Local Search1.And the proposed OLBS algorithm is compared with the online algorithm Cost First and the offline algorithm Local Search2.Both proposed algorithms are better than those algorithms in terms of total revenue and running time for scenarios with different number of edge servers and tasks.