Realistic performance evaluation of edge computing scenarios over a hybrid testbed using sofware - defined networking

Abstract

As mobile computing becomes more and more widespread along with more com plex and distributed applications, the importance of providing these applications to users at high speed and with high efficiency is increasing. One of the ways to better serve mobile users is cloud computing. However, sending data to be processed to data centers and receiving processed data take time because of the geographical separation of resources. Since users are less inclined to wait while they use their mobile applica tions, edge computing offers a better way to handle the QoS requirements for mobile users in the envisaged fifth generation (5G) mobile networks. Network maintenance and network monitoring have also a significant place if we aim to provide faster and more efficient services to users with edge computing. Therefore, an architecture which is built using OpenFlow based Software Defined Networks offers even more benefits in the context of 5G. To this end, the testbeds for Software Defined Networks are estab lished in order to test the real-life scenarios. These testbeds are usually created either using Mininet or on real hardware. Even though they have great advantages, Mininet testbeds do not produce precise results and real testbeds are not easy to expand. In this study, we aim to bring these two approaches together and provide a hybrid testbed that utilizes the features of both approaches for those who wish to perform experimentation in this field. In the experiments we have done, we presented the comparative evaluation results obtained using Mininet and our hybrid testbed for the desired edge computing scenarios. We observed different results under the same experimental variables and examined these differences with their reasons.