Keeping Data Alive: Communication Across Vehicular Micro Clouds Gurjashan Singh Pannu * , Florian Hagenauer * , Takamasa Higuchi † , Onur Altintas † and Falko Dressler * * Heinz Nixdorf Institute and Dept. of Computer Science, Paderborn University, Germany † Toyota InfoTechnology Center U.S.A., Mountain View, CA {pannu,hagenauer,dressler}@ccs-labs.org, {ta-higuchi,onur}@us.toyota-itc.com Abstract—Vehicular micro clouds are considered a prime building block for next generation Intelligent Transportation Systems (ITS) also supporting a variety of Information and Communication Systems (ICT) applications in smart cities. Such micro clouds are established by multiple cars equipped with com- munication, storage, and computational resources. We recently presented the concept of hierarchical vehicular cloud computing, which is meant to extend on Mobile Edge Computing (MEC). Based on clustering algorithms, we can set up and maintain such micro clouds and eventually make use of the distributed resources. Looking at the high mobility of cars, it is very difficult to reliably maintain data collected by cars at a given location in space that is geographically relevant, e.g., at intersections. In this paper, we propose a new protocol which encourages coordination between neighboring micro clouds to help keeping local data current, i.e., cars moving out of the micro cloud may take data to neighboring clouds, hand data over to cars moving towards the original micro cloud, and, thus, returning the data to its original geographical location. We evaluate the performance of the protocol with different vehicle densities in a Manhattan Grid scenario and our results show the benefits of our proposed inter micro cloud coordination protocol. Index Terms—Mobile Edge Computing, Vehicular Cloud, Ve- hicular Micro Cloud, Data Management I. I NTRODUCTION There has been a recent shift in focus of vehicular net- working research community towards applications supporting cooperative driving [1] and cooperative perception [2]. This is further supported as modern cars are equipped with wide range of sensors, computing, networking, and storage resources. The sensing on-board units sense abundant data from the surrounding, which can also be cooperatively used by nearby vehicles, bicyclists, and even pedestrians. Cooperative sensing can be used to maintain live 3D maps and even make complex maneuvers safely. The powerful configuration of cars makes them an important Information and Communication Systems (ICT) resource, transforming the Intelligent Transportation Systems (ITS) in future smart cities. The generated data needs to be stored and also requires frequent updates. Uploading the data to data centers is beneficial for further analysis using advanced data analytic techniques. However, vehicles interested in the data also experience longer end-to-end delays when downloading directly from data centers. To solve similar problems in cellular mobile networks, the Mobile Edge Computing (MEC) [3] architecture has been proposed. The underlying idea is to provide computing and storage capabilities at the edge of the cellular network, which is in close proximity to the users. In vehicular networks, Eltoweissy et al. [4], Gerla [5], and Dressler et al. [6] proposed the concept of vehicular cloud computing, which later evolved as the vehicular micro cloud architecture [7], [8]. In simple words, cars cooperatively form a small cluster called vehicular micro cloud which offers computing and storage services to nearby cars, pedestrians, and bicyclists, thereby, extending the concept of MEC in vehicular networks. Vehicular micro clouds can be mobile (formed by cars moving in same direction) or stationary (formed at a certain geographic region). The data stored in geographically stationary vehicular micro clouds is usually relevant to a certain geographic location. It can be uniquely identified, thus favoring communication models like Information Centric Networking (ICN) [9] and Named Data Networking (NDN) [10] rather than typical host-centric communication models. One of the main open research questions in vehicular micro cloud research is to keep the data belonging to a certain micro cloud available in the micro cloud. Each data content is associated with a unique vehicular micro cloud, which we call parent micro cloud. Our goal is to keep the data within the parent micro cloud as long as possible. This becomes very challenging due to vehicular mobility. Cars join the stationary micro cloud, collect some data, and after a while, they leave. If the car leaving the micro cloud is the last one to have certain data, then the data becomes non-recoverable in the parent micro cloud. The variation in traffic density adds more challenges on top of it. The micro cloud region can get over-crowded by cars at certain times, and within a fraction of a minute, all cars can leave the micro cloud too. In this paper, we fill in missing gaps in vehicular micro cloud research by addressing the challenge to keep data available in micro clouds. We propose a novel protocol, which encourages inter micro cloud communications. The core idea of the protocol is to allow cars in a different micro cloud to transfer the data of another micro cloud, which they brought with themselves while leaving, to those cars which will be later joining the same micro cloud. We also evaluate the protocol in many traffic densities to find the benefits of our protocol. Our contributions can be summarized as: • We introduce a novel inter micro cloud coordination protocol for stationary micro clouds, which relies upon 978-1-7281-0270-2/19/$31.00 ©2019 IEEE