Reliability has always been one of the key issues that draw widespread concerns since the rise of cloud computing. While most cloud systems use replication to ensure high reliability, some scholars incline to try erasure codes in the interest of space efficiency. Comparison of performance and reliability between replication and erasure coding have been widely analyzed by many researchers, but effort to make a thorough understanding of this issue is quite few. In this paper, we defined a quantitative model aimed at comprehensively evaluating the relative merits of the two redundancy schemes in seven metrics (availability, reliability, storage, performance, bandwidth, energy and dollar cost) which will be detailed next, and we exploited this model to provide a reference for Cloud Service Providers (CSPs for short) to set up the right configurations for their cloud file systems. In addition, we presented a scheme to decide on which one of the two redundancy schemes more suitable for a given cloud storage system.