On the efficiency of lightweight content placement heuristics for cache-enabled networks

Cache-enabled networks have received increasing attention in both wired and wireless settings. A big challenge for the operator of such networks is to solve efficiently the content placement problem, i.e., to decide how many caches to deploy in the network and in which nodes. We study the content placement problem for two classes of network optimisation objectives, the first focusing on the minimisation of the sum of the shortest paths and the second capturing the cost vs. benefit trade-off to deploy a cache. We know from the state-of-the-art that, even in small networks with few caches, it is unrealistic to find the optimal solution in a reasonable timescale for similar optimisation problems. In order to cope with this challenge, we present an approach under the prism of network analysis. We introduce a family of lightweight heuristic algorithms that use graph-theoretic metrics that identify the most important nodes of the network. We evaluate the performance of the heuristics using real network datasets, showing that the best heuristics are based on the metrics of betweenness centrality and degree centrality. Finally, we provide a randomised version of the heuristics noticing that the same metrics present again the best performance across the different datasets. Moreover, we find out that, in general, the deterministic version of each heuristic outperforms its randomised version.