Escape from Data Alcatraz
nihilist_1137 writes "Zdnet is reporting on a new information facility that is built to surive the worst.Triangular in shape, two of the sides house offices while the third, a large rectangular block if taken in isolation, contains two data centres, as well as the infrastructure to ensure that Web sites continue to function come fire, flood, natural catastrophy or foreign invasion."
"Remember thealamo.com!"
Seriously, though... you're saying they can stand up to repeated shelling by artillery? Or infantry-placed demo charges? Or anything else an invading force is likely to have?
WHY????
If you're being invaded, you've got more important things to worry about than if your company's web site will stay up!
The other half of this is: What if the invasion is an invasion of illegal immigrant workers? Can this thing survive having a janitor who's been slipped a hundred bucks (three weeks pay) to pull out a wire here and there?
By far, the cheapest and most effective method of redundant systems is to just safe your money and not buy fancy equipment for one place, but to spend it on cheap equipment is several places. That way, who cares if someone takes out an entire hosting center, leaving only a 100 ft dep crater. You still have servers running in California and Asia.
The Domain Name System doesn't rely on a huge Fort Knox-like system. It simply has 13 (?) different places throughout the world where amazingly cheap (for its importance) equipment resides. Even if North America sinks to the bottom of the Ocean, DNS should still happily resolve.
Expensive (but impressive) measures are not the answer to reliability. Geographic diversity of cheap systems is the answer most most applications. Today, we have incremental transfer protocols such as rsync that will even transfer massive databases back and forth by only sending the changes. It's largely marketing, unwarrented by technical considerations, that make companies spend so much money on these extra sigmas of reliability.