If a single component fails, a well-designed system is easily recoverable. For example, if a component failure renders the online system unusable, you should be able to replace the failed component and recover the system with no loss of data. Beware of locating stripes from different volumes on one disk—if the disk breaks, both volumes are lost.

You should consider the cost of downtime and the time required for various recovery scenarios. When downtime is expensive, you might consider more capable and reliable storage technology and extra redundancy.

Under all modes of operation, the system should not have insufficient reading or writing bandwidth through one channel and excess bandwidth through another. If you need more bandwidth, you can stripe a volume across more disks. Do not provide excess bandwidth to locations that do not need it.

Disk space requirements for some locations can only be determined by you, and depend on factors such as how large your database is.