Study On Switching And Routing Algorithms Based On Multistage Switching Network In Large-scale Data Center

With the popularity of computers and smart mobile devices,network traffic has a explosive growth while the scale of network becomes large gradually.In the network,there are many different service demands that applications such as video streaming and large-scale computation often require adequate throughput guarantee while the bandwidth demand of some applications such as email service are relatively low,that leads to a phenomenon of imbalance traffic network.Applications of big data and cloud computing migrate the storage and computing resources to the data center.Data center network adopting multistage switching structure often place the switching equipments in a distributed and long-distance way,that leads to the long distance transmission delay and path cost.Existing works in switching and routing algorithms may consider the long distance and path cost or various traffic models,however,there is little research work considering both of them.Addressed to uneven network demands to ensure 100% throughput and load balance in three-stage Clos switching network,this thesis proposes a switching and routing algorithm named Average(Heavy-Traffic First)which is based on matrix decomposition.The most significant idea is to averaging the traffic matrix to smooth the network demands and decrease the switching and routing delay.With the further study of AHF algorithm,we found that in the process of traffic matrix decomposition of AHF algorithm,there are a lot of free time slots,then we propose a switching and routing algorithm called MHS(multiple hops strategy).By shifting and transmitting packets using those empty time slots,MHS can a achieve the goal of minimizing system switching and routing delay.