Question : Fault tolerence

How much fault tolerance or redundancy is enough? When do you know enough is enough?

Answer : Fault tolerence

We used to ask "how many 9's"??  Meaning how much reliability do you need -- 99%, 99.9%, 99.99%, 99.999%, 99.9999999999999999% -- you get the idea.   The cost of each extra "9" tends to be exponentially greater than the previous one.

It's relatively easy to have sufficient fault tolerance to avoid DATA LOSS for a relatively small investment.  For example, a RAID-5 arrray, as I noted above, with a rigorous data-backup strategy can give a high degree of confidence (several 9's) that you'll not lose any data.  But this is different than avoiding DOWNTIME.  Systems can be built so they don't lose data when they fail -- but if you need to avoid downtime than you need redundant SYSTEMs as well.  This is one reason virtually all mission critical organizations have UPS, generators, and in some cases have even built their own fuel-cell power plants, in addition to built-in redundancy in all their data processing.   That's also why "big iron" still rules in the mission-critical banking, insurance, utility, etc. businesses.  The big mainframes have ALL of these characteristics engrained in their design.