In large-scale distributed systems such as data grids or cloud computing environments, the efficacy of replication strategies is significantly affected by the location of replicas. The geographic location of data is a critical factor in the optimal placement of data replicas. It directly impacts data availability and response times within large-scale distributed systems. Data access models, characterised by various locality properties (temporal, geographical, spatial), considerably influence system performance. Most proposed replication strategies do not take into account these complex properties of access models, which has led to suboptimal performance. We propose in this paper a new replica placement approach based on neighbourhood and that exploits these locality properties in order to optimise the performance of large-scale distributed systems. Simulation results show a significant improvement of 18% in response times and show that the replication frequency was significantly reduced by 40%.