..for the end of the MTTF paper document -------- - Carl ====================================================================== \pagebreak \section{Examples} Let $\lambda = 0.5$ failures/year, assuming a node fails on the average once every two years. Let $\mu = 52.0$ repairs/year, assuming a node can be repaired in a week. These assumptions result in the following system MTTF values, in units of years: \begin{tabular}{||r|r||r||} \hline $N$ & $K$ & MTTF \\ \hline 2 & 1 & 107.00 \\ 3 & 1 & 3747.67 \\ 3 & 2 & 36.33 \\ 4 & 1 & 98405.50 \\ 4 & 2 & 955.50 \\ 4 & 3 & 18.50 \\ \hline \end{tabular} If $\lambda = 0.5$ failures/year and $\mu = 350.0$ repairs/year (to follow the Stratus practice) we get the following system MTTF values, in units of years: \begin{tabular}{||r|r||r||} \hline $N$ & $K$ & MTTF \\ \hline 2 & 1 & 703.00 \\ 3 & 1 & 164270.33 \\ 3 & 2 & 235.00 \\ 4 & 1 & 28788554.16 \\ 4 & 2 & 41185.50 \\ 4 & 3 & 117.83 \\ \hline \end{tabular} One could also consider the MTTF of a posting on USENET News. $\lambda$ and $\mu$ might be the same as above, but $N$ is the number of News servers to which the posting has propagated and $K$ is 1. The underlying model is different because it must take account of limited connectivity of News servers, but it is clear that a posting's MTTF becomes effectively infinite once it has left its home node. \end{document}