Traffic-Profile and Machine Learning Based Regional Data Center Design and Operation for 5G Network

Udita Paul, Jiamo Liu, Sebastian Troia, Olabisi Falowo, and Guido Maier

105 569.569-582


Abstract :Data center in the fifth generation (5G) network will serveas a facilitator to move the wireless communication industry froma proprietary hardware based approach to a more software orientedenvironment. Techniques such as Software defined networking(SDN) and network function virtualization (NFV) would beable to deploy network functionalities such as service and packetgateways as software. These virtual functionalities however wouldrequire computational power from data centers. Therefore, thesedata centers need to be properly placed and carefully designedbased on the volume of traffic they are meant to serve. In this work,we first divide the city of Milan, Italy into different zones using Kmeansclustering algorithm. We then analyse the traffic profiles ofthese zones in the city using a network operator’s Open Big Dataset. We identify the optimal placement of data centers as a facilitylocation problem and propose the use of Weiszfeld’s algorithmto solve it. Furthermore, based on our analysis of traffic profilesin different zones, we heuristically determine the ideal dimensionof the data center in each zone. Additionally, to aid operation andfacilitate dynamic utilization of data center resources, we use thestate of the art recurrent neural network models to predict thefuture traffic demands according to past demand profiles of eacharea.​ 

Index terms :Big data, cellular traffic, data centers, recurrent neural networks, traffic prediction, 5G.