Looking at the Reliability Calculator on Jason Johnson's setup (1.9.2), and
then taking all defaults except for the last one (report_span), which I set to
25Y from the default 5Y, I noticed that the probability of failure without
maintenance peaks at 15 years and 7 months at about 0.628, and then starts
decreasing again, which seems extremely odd.
Then changing drive lifetime to 7Y from 8Y, the peak changes to 13 years 6
months, also just over 0.628.
Then changing N to 15, the peak changes to 16 years 3 months, at 0.615
Then changing k to 4, the peak changes to 14 years 4 months, at 0.760
Going back to the defaults, except for changing k to 4, gives us a peak of 13
years 2 months, at 0.777.
Hopefully this will help track down the issue with the math (assuming it wasn't
fixed in 1.10.0 - I'm still working through python basics to be able to quickly
build test grids).
_______________________________________________
tahoe-dev mailing list
[email protected]
https://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev
