Reliability Design for Large Scale Data Warehouses

Data reliability has been drawn much concern in large-scale data warehouses with 1PB or more data. It highly depends on many inter-dependent system parameters, such as the replica placement policies, number of nodes and so on. Previous work has roughly and separately discussed the individual impacts of these parameters, and seldom provided their optimal values, nor mentioned their optimal combination. In this paper, we present a new object-based-repairing Markov model. Based on analyzing this model in three popular replica placement policies, we figure out the individual optimal values of these parameters at first, and then work out their optimal combination by GA. Compared with the existing models, our model is easier to solve while reaching more integrative and practical conclusions. These conclusions can effectively instruct the designers to build more reliable large-scale data warehouses.