DHFogSim: Smart Real-Time Traffic Management Framework for Fog Computing Systems

Abstract

Clouds are the most powerful computation architecture; nevertheless, some applications are delay sensitive and need real time responses. Offloading tasks from user device to the cloud will take relatively long time and consumes network bandwidth. This motivates the appearance of fog computing. In fog, computing additional layer falls between user device layer and the cloud. Offloading tasks to fog layer will be faster and save network bandwidth. Fog computing has spread widely, but it is difficult to build and test such systems in real word. This led the developers to use fog simulation frameworks to simulate and test their own systems. In this paper, we adopt fog simulation formwork, which adds smart agent layer between user device and fog layer. The framework uses multilevel queue instead of single queue at the Ethernet layer, these queues are scheduled according to weighted round robin and tasks dispatched to theses queues according to the value of Type of Service (ToS) bits which falls at the second byte inside the IP header. The value of ToS bits given by the smart agent layer according to take constraints. Framework behavior compared with mFogSim framework and the results shows that the proposed framework has significantly decrease the delay on both brokers and fog nodes. furthermore, packet drop count and packet error rate are slightly improved