Data access latency is one of the important factors of system performance in federated data grid. However, different replication strategies resulted in different access latency. This paper proposes a model of replication strategy in federated data grid that is known as dynamic replication for federation (DRF). DRF uses the concept of a defined ‘network core area’ (NCA) as the designated search area of which a search will be focussed. However, the area known as NCA is not static and it is bound to change if data requested cannot be found. This paper will also highlight how NCA is defined and reallocated if a search fails. Results from the analysis show that the DRF proposed in this paper is superior to Optimal Downloading Replication Strategy (ODRS) in terms of wide area network bandwidth requirement and in the average access latency data where the results was 48% higher than for
=0.3, 0.4, 0.