A two-step chunk-based algorithm for offloading streaming traffic through a vehicular cloud

Using vehicles equipped with small caches as small base stations has recently been proposed as an interesting middle ground between caching at fixed base stations (which has higher CAPEX/OPEX), and caching at user devices (which has resource limitations). A typical problem in this setup is which content to store in which vehicles. The correct answer depends on the application. Indeed, if the stored content will be streamed (not downloaded), then this offers a natural delay tolerance: latter parts of the content do not need to be downloaded immediately from expensive links (e.g., macro-cells), but could be fetched from encountered vehicles cheaply. In an earlier work, we formulated a related optimal cache allocation problem, in which the proposed solution stores a content in its entirety. In light of recent statistics suggesting that different parts of a content (e.g., YouTube clips) are not watched equally frequently, this method is suboptimal. In this paper, we thus consider per-chunk allocation, and propose a simple two-step heuristic that first allocates the vehicular cloud capacity among content items, then efficiently distributes the capacity for a specific content among its chunks. Trace-driven simulation results suggest that chunk-based allocation can lead to considerable gains.