Design And Implementation Of Inter-DC Traffic Scheduling System Based On Prediction

To handle the problem of low link utilization resulting from mixing online and offline traffic in one data center transmission network and separating them with a fix way in the same link,we propose a solution of offline traffic scheduling based on online traffic prediction.At the same time,an effective failure recovery mechanism is used to ensure the availability of traffic demand.Experimental results show that the combination of Sliding-k,SEDF and incremental fast rerouting can improve the utilization of data center links and ensure traffic availability.The main research work of the article includes three parts:prediction,scheduling,and failure recovery.Firstly,prediction.It predicts online traffic needed to be guaranteed preferentially in link using a method calling Sliding-k that combining exponentially weighted moving-average(EWMA)algorithm and bayesian changepoint detection algorithm.It makes prediction sensitive to a suddenly change of network environment and reduces unnecessary re-adjustments when network environment is steady at the same time.Secondary,scheduling.Then,it computes the remaining space for offline traffic according to online traffic prediction result and implements dynamic bandwidth allocation.After that,it uses an algorithm called smart early deadline first(SEDF)that can consider both traffic deadline and size to schedule offline traffic.Thirdly,failure recovery.Finally,when a link failure occurs,a method based on the Value at Risk theory is used to quickly reroute the traffic on the failed link,so that the traffic demand that was originally scheduled to the failed link can be restored.In addition,the article developed a inter-DC traffic scheduling system based on prediction(IDCTSSP)based on the above three research work.The system realizes the functions of online traffic prediction,offline traffic scheduling,and recovery in the event of a failure,and integrates these three functions into an overall traffic scheduling system.In addition,the system also has common functions of the system such as user login and registration,system parameter management,and scheduling results viewing.This thesis first introduces the research background,objectives and content of this work,analyzes the situation of inter-DC traffic scheduling work,explains the necessity of improving link utilization and guaranteeing the availability of offline traffic and then introduces the main technology used in this system.On this basis,it makes the demand analysis and designs,clarifies what problems and challenges exist in the current scenario of inter-DC traffic scheduling.It gives the overall architecture design of the system based on demand analysis,and then refines the system function modules into prediction modules,scheduling modules and fault recovery modules,and describes the detailed design and implementation of each module.Then it describes the testing of the system including white-box testing and black-box testing,which verifies the usability and stability of the system,and finally makes a summary of the full thesis.