It seems now that everything is going slowly back to normal for Facebook services! Now let’s see how the “who’s going to be guilty of this” game follows..
Indeed, but I find this to be... demonstrative, for sure. Facebook is supposed to operate at high-tier 6-sigma, and connects to a network of 6-sigma service providers. It works out to about 1m47s of downtime a year. This is achieved by multiple layers of redundancy etc, so that every component can fail flawlessly without affecting the effectiveness of the system. RAID drive arrays, alternate server towers lending a shoulder under load or distributed mirrors switching in when the primaries go down... at every step, it's incredibly thought through, and every time I talk to my brother (systems architect working remotely for a big-tech related firm) I realize that they've already thought 3 steps ahead of the challenges I underline.
So a 6-sigma company having an outage for 6 hours in a single day (that is already down to 4-sigma reliability, if the rest of the year goes well) is very meaningful. It's probably confirmation bias if I said it's a sign of civilization collapsing if we cannot maintain a modern level of reliability. But it seems unlikely to me that a collapse of reliability would happen without many of the partner service providers activating various failsafes and mirrors. How did it happen? Might be a huge coincidence. But that would be very unlikely. So, by Occam's...
Sabotage? Would be better than the internet collapsing in reliability. I don't care much about turf wars between psychopaths.
Edit: Jasophoris, I just read your message - if FB is still partially/mostly down, that's longer than 6 hours! We're talking about its service tier falling down lower into the 4-sigmas... interesting!
Re-edit, I'm late to the party, ZH had this to say earlier today:
To be more precise (and Geek Factor 5) the BGP routes serving Facebook's authoritative DNS were withdrawn, rendering all Facebook domains inaccessible. That's per @DougMadory , who knows a few things about BGP/DNS.
ZeroHedge - On a long enough timeline, the survival rate for everyone drops to zero
www.zerohedge.com