The reliability of RAID storage systems can be increased according to the data redundancy method, which uses multiple disks, and parts of these disks are used to store redundant information, when disk failure occurs, system can utilize redundant information to recover RAID systems; however this will lead to more time to recover data and reduce the storage performances. Even worse, during data rebuilding, the possibility of secondary disk failure will increase. This paper proposes a new RAID-based system, which is defined as REA: RAID storage system with Early-warning and Active data migration. REA can predict RAID systems potential faults according to the disk health degree model. Once the fault has been predicted, REA uses active data migration technology to protect the data. REA can effectively avoid the complex checksum computation of the rebuild process and dynamically adjust the speed of data migration according to the system I/O load in order to reduce the RAID systems’ performance. The overall results indicate that REA can achieve very high prediction accuracy and improve performance and reliability of storage systems with few side effects. Numerical results using SPC traces and standard benchmarks show that REA can effectively improve RAID reliability and performance.