Decentralized Algorithm for Joint Data Placement and Rate Allocation in Content-Aware Networks

In this paper a mathematical model of data delivery in Content-Aware Network (CAN) is studied. Such network architecture assumes that users access data objects (content) replicated on a number of cache servers. A CAN operator decides on the placement of data objects on behalf of the content publishers, and decides on the transmission rate allocation of users (content consumers downloading content, and publishers uploading and updating their content). A problem of minimizing the mean transmission delay is considered and formulated as a mixed-integer programming problem. Decomposition-based heuristic is proposed, which utilizes projected gradient descent and greedy strategy. The algorithm is presented as a decentralized protocol, embedded in the network architecture.