Latency-optimizing file splitting for transmission over a large multi-hop network

We consider the problem of latency-optimizing file splitting for transmission over a large multi-hop network. We utilize recent results from random matrix theory to analytically express the latency distribution as a function of the relevant network parameters and the file size. The resulting minimax problem can be solved using standard nonlinear programming techniques for a variety of latency metrics. We illustrate the reduction in latency obtained using our technique, relative to some commonly-used heuristics, in the large-file-large-multi-hop network setting and highlight the role of random matrix theory in realizing the resulting performance gains.