This document discusses a fusion method for tolerating faults in distributed systems. The fusion method uses a combination of error correcting codes and selective replication to tolerate crash faults. It maintains "K" additional fused backups instead of "K+1" replicas required by replication. This provides an O(n) space savings over replication and results in minimal overhead during normal operation. The method encodes primary data structures into the fused backups. It can insert, delete, and update data in both the primary structures and backups efficiently. The proposed fusion technique saves space for fault tolerance compared to replication.