In large-scale cloud data centers, correlated failures in which a single fault simultaneously disrupts multiple physical machines pose a severe threat to job reliability. The DieHard framework of \citet{sedaghat2016diehard} addresses this problem by modelling power and network failure domains and providing a fault-aware scheduling algorithm that achieves a desired reliability threshold with the minimum number of redundant tasks. The DieHard framework did not account for dependence among power failure domains when calculating job reliability. However, real-world power infrastructures are subject to common cause events such as regional power-grid disturbances or simultaneous PDU failures triggered by a shared electrical fault, which induce statistical dependence among power domains. This thesis proposes an extension of the DieHard reliability model that adjusts the independent power failure domain probability with a dependence model, in which correlated power failure domain (PFD) failures arise from three power distribution units (PDU) lifetimes shared across the power failure domains. We incorporate the modified joint probability mass function into the overall job reliability expression. In addition, we adapt the DieHard fault-aware scheduling algorithm to the revised model. A simulation study implemented in R and inspired by the experimental configuration of \citet{sedaghat2016diehard} evaluates the modified model and two other commonly used placement strategies across a range of different scenarios. The results demonstrate that no single placement algorithm is the best across all scenarios. The preferred placement algorithm depends on the correlated failure structure of the data centre. It is also observed that the proposed adjusted DieHard model always works equally well or better than the unadjusted DieHard model.