A Study Of Multi-path Scheduling Algorithm In SDN

With the new technologies of big data and cloud computing used in people's daily life,applications of network transformed simple way into interaction are increasingly deployed on the huge data center.Network business in corresponding data center shows new complex and diverse characteristic,therefore,network traffic will be increased rapidly,especially the serious problems of network congestion which are caused by the huge “horizontal” traffic.Regarded as representative of the new network paradigm,Software defined network(SDN)provides excellent methods to solve the problems of network congestion which are lead by sudden large flows of data in data center network.Using SDN architecture based on Ant Colony algorithm with elite,this paper has been researched on multi-path scheduling with large flows of data.Mainly studying aspects as following.Firstly,owing to much more the number of nodes at data center network,the problems of slow convergence would be emerged if we only use Ant Colony algorithm.However,if we optimize Ant Colony algorithm with elitist strategy strongly retaining the current optimal value in the update process of information table,the experiments would be converged quickly in the better range of results.Secondly,our research has studied problems of multi-path scheduling using Ant Colony algorithm with elite.In order to dealing with the congestion problems caused by unexpected data on traditional single optimal path,this paper mainly utilizes the technologies of multi-path scheduling and reasonably use other free links to resolve single-path congestion.We take several measures to dynamically dispatch the flow of data.For example,the table of information and data flows have the same structure,link-free rate is appointed as transferring probability to select the direction of route,ants proactively monitor links in the process of looking for direction of links,and the live information of QoS is updating the table of information and data flows.Finally,based on Linux platform,simulation software of Mininet was combined with Floodlight controllers to build network simulation environment of SDN.During the experiments,the topology of network was designed by Python language,and series of experimental data was obtained by Iperf flow tools.The results of the experiment indicated that compared with the phenomena that static algorithm of ECMP learned overlaps equivalent multipath and SPF algorithm got optimal single path,Ant Colony algorithm with elite would receive several scheduling paths with different direction.Therefore,Ant Colony algorithm with elite proposed in this paper made full use of the energy of network idle links,and improve the problems of network traffic caused by single scheduling of large flows of data.