Answers are grounded in this paper and cite their sources.
There is danger in coming to important conclusions based on averages, as insights into spread are not provided by this single measure. When promises are made about availability this is a "mean". Knowledge of availability probabilistic distribution is paramount in understanding what is being offered and what is being bought, as the chance an availability SLA will be violated can be ascertained. Here, the distribution of a typical availability promise, "Five-Nines" or 0.99999, is determined for different DS-3 reliability and maintainability scenarios. As two random variables are in play, "time to fail" and "time to restore", the determination of availability distribution is analytically intractable. Monte Carlo simulation, with exponential failure and repair distributions, is performed to gain insights into availability distribution. The principal findings are that significant chances Five-Nines availability will be violated exist, and that the reliability/maintainability strategy used by the carrier determines availability SLA violation risk.
No takes yet. Share an insight, caveat, or question.
Snow et al. (2007) studied this question.