Availability via Redundancy and Diversity

As argued, while sooner or later a rolled die will come up with a six or any other number we select, the chances that all rolled dice land on a given number become lower as the number of rolls increases. It is relatively unlikely (1 chance out of 216) that the first die lands on six and that the second die lands on six and that the third die lands on six. If, instead of dice, we are considering data center outages, it means that the chance that the first data center is down and that the second data center is down and that the third data center is down is extremely unlikely, assuming that outages are independent.

The chance that the original instance of the data is lost in a disaster and that a remote copy of the data is lost in a disaster and that another copy of the data in a third location is lost in a disaster is also extremely unlikely, again, assuming independent events, such as hurricanes and fires.

Merely increasing the number of components, as we argued, can reduce availability. But if those components are appropriately replicated, with cross connections between parts of the system to ensure functionality even in the event of a single point of failure, the overall availability can increase.

Availability tends to increase with redundancy. If a single component has an inherent availability or reliability of 90%, then two of them have 99%, three have 99.9%, and four have 99.99%. If the single component has an inherent reliability of ...

Get Cloudonomics: The Business Value of Cloud Computing, + Website now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.