Everyone wants storage arrays that keep on working no matter what. You'd think that the verbal description vendors tack onto their products would indicate whether the product meets that goal or not. But it seems that differences in terminology can indicate important differences in reliability.
The RAID Advisory Board (RAB), a trade group for the RAID industry, has established three major classes of protection (with an additional four subvariants) to identify the amount of reliability RAID arrays offer.
"Failure Resistant" is RAB's name for a RAID array that keeps reliable on-line data available immediately in the event of the failure of one disk in the array. In addition, the contents of the failed disk are reconstructed on-line and recorded on a replacement disk while maintaining an acceptable level of I/O performance. If the disks, controllers or other Field Replaceable Unit (FRU) in the storage system fails -- except the storage cache -- the data is protected.
"Failure Tolerant" systems meet those criteria and several others as well. For one thing, the data is protected if the storage cache fails. Data is also protected in the event of a power failure or overheating, the system supports disk hot-swapping and on-line spares and the data is available immediately in the event of the failure of any one storage system FRU, including the cache and power supply.
A "Disaster Tolerant" system does all those things and also supports multiple I/O channels and multiple primary power sources. All FRUs are hot-swappable and the data remains available in the event of failure of any one zone in a multi-zoned storage system which provides physical separation of the zones.
A more comprehensive discussion of all seven classes of RAID systems is available at the RAB web site.
Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.