This paper introduces a general framework for analyzing the bandwidth requirements of storage architectures that combine coding and replication in distributed storage systems. The authors present two new schemes for maintaining redundancy using erasure codes: Optimally Maintained MDS (OMMDS) and Regenerating Codes (RC). OMMDS optimizes the generation of MDS fragments directly from existing fragments, while RC uses slightly larger fragments than MDS but requires less overall bandwidth. Through simulations, the authors show that RC can reduce maintenance bandwidth by up to 25% compared to the best previous design (Hybrid), a hybrid of replication and erasure codes, while simplifying system architecture. The paper also discusses the trade-offs between RC and other strategies, emphasizing that RC offers a promising alternative due to its simplicity and low maintenance bandwidth.This paper introduces a general framework for analyzing the bandwidth requirements of storage architectures that combine coding and replication in distributed storage systems. The authors present two new schemes for maintaining redundancy using erasure codes: Optimally Maintained MDS (OMMDS) and Regenerating Codes (RC). OMMDS optimizes the generation of MDS fragments directly from existing fragments, while RC uses slightly larger fragments than MDS but requires less overall bandwidth. Through simulations, the authors show that RC can reduce maintenance bandwidth by up to 25% compared to the best previous design (Hybrid), a hybrid of replication and erasure codes, while simplifying system architecture. The paper also discusses the trade-offs between RC and other strategies, emphasizing that RC offers a promising alternative due to its simplicity and low maintenance bandwidth.