Experimental evaluation of fog computing techniques to reduce latency in LTE networks

Abstract

Abstract The changes in new mobile networks toward a full Internet protocol–based architecture have led to opportunities for service‐oriented optimizations based on emergent technologies like fog computing, software‐defined networking, or network function virtualization. This paper explores 2 ways of using these new technologies to reduce the latency in Long‐Term Evolution (LTE) networks. Both solutions reduce the path that the data packets should follow from the base station (evolved Node B [eNB]) to the network components that connect to the servers. The first solution, called Fog Gateway, is based on the interception of the packets in the tunnel at the eNB and their redirection to local servers running the fog services. This solution is fully compliant with the current LTE architecture and only requires new components. The second solution, called General Packet Radio Service Tunneling Protocol Gateway (GTP), is based on splitting the eNB's functionality to avoid unnecessary GTP encapsulation of the packets geared toward the fog services. This paper includes an analysis of the latency split in LTE networks, the evaluation of both solutions with experiments in an end‐to‐end LTE network testbed, and a discussion around their applicability in future fifth‐generation networks. The results confirm that they are feasible to provide low‐latency services and that they are compatible with some of the emergent paradigms (software‐defined networking and network function virtualization) as well as with the studies on fifth‐generation networks from the standardization bodies.