Research On Bandwidth Scheduling Algorithm Based On Dynamic Pricing In High Performance Networks

With the rapid development of information technology,massive amounts of data have been produced in the deep integration process of traditional industries and emerging Internet economy.The development and growth of big data industry has been directly promoted by the development of storage technology and computing technology.The volume of data grow rapidly,giving rise to the proliferation of cloud computing and data centers around the world.Connecting backbone networks between remote data centers is an onerous task for data transfer.Cloud service providers use expensive,high-bandwidth links to deliver big data services to a variety of users.How to earn considerable profits by providing high quality services is the core issue that service providers pay attention to.Scientific and reasonable pricing mechanism is an important link to ensure operation.At present,all the major cloud service providers adopt relatively simple pricing methods based on different regions and different periods,such as Amazon and Google.And there is no guarantee that the user's transfer request will be completed before the deadline.This relatively simple pricing method can not adapt to the user's demand for different transmission service levels,which has some shortcomings for both users and service providers.In this paper,we study a dynamic pricing bandwidth schedule in the high performance backbone network between data centers that exploring a dynamic pricing strategy to provide appropriate service priorities for batch user big data transmission requests,and allocating path bandwidth resources for different user requests based on the priority:(1)User request interface module design.The operator will generate a bandwidth price list p(·)and the upper limit of available bandwidth capacity bt according to the current state of the network and the request information when the user request arrives.p(b‘)in the price list p(·)is the price corresponding to bt of the transmission bandwidth guaranteed before the deadline.bt is the maximum bandwidth capacity that the operator guarantees to route before the cut-off date.Each link has a certain price at each time slot,and the pricelist p(·)first routes along the least-price path until the path is saturated,then looks for the next least-price path,and the available bandwidth of each path is priced accordingly,and so on.Since the price of each available path only increases,the bandwidth pricelist p(·)is a non-decreasing,convex piecewise linear function.(2)Conduct dynamic pricing for each time step of each link.The new link utilization rate and the link price data pair of the previous time slot after the link utilization rate changes continuously from 0.1 to 0.8 are recorded.The least square nonlinear fitting method is adopted to fit the above data,and the link utilization rate and price function model of the real operation state is obtained.The link utility-price function is updated every time 20 consecutive sets of new link utility-price data(link utilization in the range of 0.1-0.8)are collected.After the link utility-price function is obtained,the current price of each link is calculated by the nonlinear fitting function using the current link utilization.For the bandwidth reserved for high-priority requests with a link utilization rate exceeding 0.8,the link price is set to 2 times the price corresponding to the link utilization rate of 0.8.(3)The Dynamic pricing scheduling algorithm and Region-based pricing scheduling algorithm are simulated.Using 10 different numbers of requests for 50-500 requests,test the network efficiency achieved with the two pricing scheduling approaches.In addition,the operator's profit,the completion rate of the request,the network utilization rate and the sensitivity are also analyzed.