Lessons From Recent Power Outages

Mike O'Dell mo at 131.ccr.org
Thu Jul 19 15:35:59 CDT 2012

i agree completely.

once you reach a certain scale,
the ONLY way to achieve availability and reliability
is by way of the architecture of the applications
because the simple brute-force of making the platform
sufficiently reliable does not work no matter how much
money you spend. (and you will likely wish to stop
spending money long before abject impossibility)

this is an example of "Non-Newtonian Computing".
speed cannot be infinite,
delay cannot be zero,
and rigid synchronization is pathological.

oh yes - grief is conserved, mostly.


More information about the Tacos mailing list