Abstract: m of n fault tolerance techniques are used to increase availability and reliability in many P2P network storage systems, but in practice correlated failures among servers will lead to low tolerance rate of those techniques. Presented in this paper is an approach to discover sets of server nodes that fail with low correlation. m of n fault tolerance techniques can be increased by using these independently failing server nodes. The approach is discussed in detail and a preliminary evaluation is provided, which shows that the approach is practical and effective.