Efficient replication for vehicular content distribution

We consider a vehicular content replication system which makes use of the deployed Access Points (APs) to maximize the vehicular download progress of delay-tolerant contents through replication in the APs' local storage. The replication system for vehicular users is quite different from the traditional system for static Web users. The transient connection period between the vehicle and the AP makes it difficult for the vehicle to download the entire file requested and thus the content retrieval is usually across several APs. Such characteristic poses two problems: (1) replication in units of entire file may be inefficient in terms of resource utilization; (2) the actual contribution of an individual AP can be affected by the other correlated APs during the content retrieval. To deal with these challenges, we formulate the vehicular content retrieval into an offline optimization model that helps establish the performance bounds of replication algorithms in maximizing vehicular download progress. A real vehicular trace is also thoroughly analyzed. Then we propose an efficient and distributed replication algorithm explicitly taking into account the content popularity, vehicle-AP contact pattern and content availability among correlated APs. Simulation based on real vehicular trace proves the effectiveness of the proposed replication system. The performance in terms of download rate and completion ratio has at least 15% to 20% improvement against the algorithms under comparison.