GW elasticity in data centers: Options to adapt to changing traffic profiles in control and user plane

This paper discusses different implementation options of telecommunication network GW functions like the 3GPP Evolved Packet Core (EPC) S/PGW in Cloud computing environments. It is focused on the question what benefits can be achieved if the computing resources can adapt to fluctuating traffic demands. In particular also the traffic profile may change what may lead to different GW control plane and user plane demands. Two implementation options are compared: one running a complete GW in a virtual machine and the other decomposing the GW into a control and user plane virtual machine. For this investigation a traffic model is introduced that assumes not only load changes between busy hour and times with little network load but also a different application profile during the time periods. Several example calculations give an indication on the sensitivity of different parameters to the scalability of the implementation options and resource savings.