3 ms·
I recall that Google has a library that will answer the question, "given two identifiers like cluster names or machine names, which failure domains do they shar
by philsnow 2y ago
I recall that Google has a library that will answer the question, "given two identifiers like cluster names or machine names, which failure domains do they share?". Those failure domains can be like "they're in the same rack", "they share a power bus", "they're in the same physical building", "they're in the same geographical region", that kind of thing.
I wonder if it has since been updated to account for different disaster scenarios. It would be good to know, if you're storing 5 redundant copies of something critical, that all 5 copies won't be under 10 feet of water if an asteroid strikes one of the poles and melts all that pole's ice.
I can't imagine what data would be so critical that it's important to make sure it survives such an event, but that hasn't stopped googlers over-engineering internal things in the past.
- pigscantfly 2y agoThe LOCKSS [1] digital preservation threat model (I've been told, there's a whitepaper if curious) considers nuclear first strikes and large solar storms -- they maintain a global network of servers for archiving and preserving academic research. [1] https://www.lockss.org/why-lockss https://www.lockss.org/why-lockss