ABSTRACT
One of the significant research challenges in the realm of vehicular clouds is to identify conditions under which these clouds can support big data applications. It is clear that big data applications, with stringent data-processing requirements, cannot be supported by ephemeral vehicular clouds, where the residency time of vehicles in the cloud is too short for supporting virtual machine setup and migration. Similarly, it turns out that vehicular cloud implementations relying on bandwidth-constricted interconnection topologies are not suitable for big data applications. Unfortunately, this is the case of the vast majority of vehicular clouds proposed thus far in the literature that rely on a wireless interconnection fabric.
Our main contribution is to identify sufficient conditions under which big data applications can be effectively supported by datacenters built on top of vehicles in a parking lot. This is pioneering work: to the best of our knowledge, this is the first time researchers are looking at evaluating the feasibility of the vehicular cloud concept and its suitability for supporting big data applications. One of our main findings is that (1) if the residency times of the vehicles are sufficiently long and (2) if the interconnection fabric has a sufficient amount of bandwidth, then big data applications can be supported effectively by such datacenters.
