MIST: Mobility-Inspired SofTware-Defined Fog System Haymanot Gebre-Amlak * , Seoungjin Lee * , Abdoh M. A. Jabbari * , Yu Chen † , Baek-Young Choi * , Chin-Tser Huang ‡ , and Sejun Song * * University of Missouri - Kansas City Email: {hhgc77, seoungjin.lee, jabbaria, choiby, songsej}@umkc.edu † Binghamton University, SUNY, Email: ychen@binghamton.edu ‡ University of South Carolina, Email: HUANGCT@cse.sc.edu Abstract—Softwarization approaches in networks, storages, M2M, services, and smart things aim to optimize costs and processes and bring new infrastructure definitions and functional values. A recent integration of wireless and mobile cyber physical systems with the dramatically growing smart sensors enables a new type of pervasive smart and mobile urban surveillance infrastructures, which opens up new opportunities for boosting the accuracy, efficiency, and productivity of uninterrupted target tracking and situation awareness. In this paper, we present a design and prototype of a mobility- inspired efficient and effective fog system using software-defined control over a mobile and wireless environment (MIST). Fog Computing, a recently proposed extension and complement for cloud computing, enables computing at the network edge in a smart device without outsourcing jobs to a remote cloud. We investigated an effective softwarization approach in the Fog environment for dynamic big data driven, real-time urban surveillance tasks of uninterrupted target tracking. We address key technical challenges of node mobility to improve the system awareness. We have built a preliminary proof-of-concept MIST architecture on both Android and Linux based smart devices and tested various collaboration scenarios among the mobile sensors. I. I NTRODUCTION An array of traditional urban surveillance infrastructures, such as stationary video cameras and RFID sensors becomes smarter when they are connected to central control systems by networks where Cloud services and applications can store, analyze, and control the collected data to help make various decisions . Furthermore, a recent integration of wireless and mobile computing technologies with the dramatic growth of smart devices enables a new type of crowd-based pervasive smart and mobile urban surveillance infrastructures. This, which opens up new opportunities for boosting the accuracy, efficiency, and productivity of uninterrupted target tracking and situation awareness. In order to ensure effective data col- lection, efficient information abstraction, and instant decision making, a Fog computing [23] is used at the network edge that keeps data and their processing among heterogeneous smart mobile devices. However, a few challenging issues of building a smart device based fog computing system includes the device mobility and service heterogeneity controls. Figure 1 illustrates the system architecture level vision of Fog computing paradigm for Big Data Driven, real-time Information, Surveillance, Target Acquisition, and Recon- naissance (ISTAR) [8] in an urban response area. Multiple Fig. 1. Urban Surveillance Architecture sensor units, including satellites, UAVs [15], mobile robots, vehicles, social networks, and first responders are monitoring the area concurrently from different positions. When they are collecting real-time data streams, each of them also needs processed information for instant decision making. Although it is ideal that all the collected data are sent back to the central Cloud facility for thorough global analysis, there is no guarantee that a reliable communication network to a remote Cloud center is always available. In addition, not all data are globally significant, do not create necessary traffic in the networks. Instant on-site decision making also reduces the risk of exposing the data to eavesdroppers in transmission channels. Just like a powerful centralized processing center, a well-equipped Cloud is separated by an area of uncertainties. It is not agile enough to handle latency and connectivity sensitive tasks as well as cannot guarantee reliable and secure communication services for many mission critical situations. Therefore, the Fog, which consists of the computing devices carried by the units in or near the disaster area, can fulfill the requirements very well. Live event processing necessitates a service-oriented, dynamic data driving work model. Modern virtualization technology enables the Fog to provide a uniform task-oriented, homogeneous computing platform on top of the