Optical networking for hybrid computing combining cloud and fog
Yongli Zhao, Xinbo Wang, Jie Zhang, Biswanath Mukherjee · 2016
Driven by ubiquitous cloud computing application, datacenters have been deployed widely as important IT resources. However, some disadvantages are challenging the cloud computing because the datacenters are deployed concentrated and remotely, which cannot guarantee the quality of experience (QoE) for the customers, such as the latency, bandwidth, energy consumption, and network cost. Fog computing has been proposed by Cisco in 2011, the goal of which is to improve efficiency and reduce the amount of data to be transported to the cloud for data processing, analysis and storage. Compared with cloud computing scenario as shown in Fig. 1, the traces of traffic flows in fog computing scenario will change to be locked locally to reduce latency and save network resource for some applications, such as online games and online searching. However, fog computing cannot replace cloud computing in the future, because the latter also has some obvious advantages, such as security, sharing, and scalability, which means that cloud computing and fog computing will be co-existing in the future. The concept of hybrid computing is proposed in the paper, considering cloud and fog together, as shown in Fig. 2. Then, two kinds of problems will emerge in hybrid computing scenario, one of which is how to deploy fog computing resource at the network edge to cooperate with cloud efficiently, and the other of which to how to provide network resources for hybrid computing scenario because the traffic flows will be divided into different tiers according to their requirement and the status of IT resource and network resource. Focusing on the problems above, the authors will give an optimized deployment strategy of fog computing resources, and then will proposed a software defined optical networking architecture for hybrid computing considering the service characteristics. Some numeric results and analysis will be given finally.