Off-site redundancy is truly the way to go. Since a database of files are replicated accross multiple server stations, fault tolerance is less of an issue concerning file integrity (free of errors).
CRC (Cyclic Redundancy Checks) assures that data is mirrored with the same exact file so only integral files are mirrored accross the cloud servers. This prevents errors from being duplicated accross the servers.
With RAID data is mirrored among an array of hard disks WITHIN a computer server. This does nothing if the errors transfer from disk to disk!
A good online backup will not DUPLICATE errors! Backup servers are not just for mirroring, but to assure every file is intact and without errors.
Definitions:
Data Replication
Data Redundancy
Amazon Cloud EC2 or S3 and Online Cloud Backup Systems are totally different.
The Amazon Cloud is mainly a "hard-drive in the sky" to work applications into a 'bucket' dump files area on their servers. Large amounts of data can be processed on the Amazon S3 hard drive for online based applications such as Digital Goods processing.
The Amazon Cloud (or others like them such as Yahoo! or Google products) are not backup systems. It is the same as using one's own server and hard drive for applications and storage computing. These provide larger storage capacity 'GB or TB' to run applications.
Amazon states in their Terms of Agreement that Amazon is on an "As is" basis with no guarantee or warranty. Accounts may be terminated at any time and Amazon claims no responsibility. Data can be lost without a restore guarantee. Amazon has gone blank before.
A company using the Amazon Cloud should consider using a 'cloud backup system' to backup data being sent to the Amazon cloud. Please contact me for questions if this could work for your application.