Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I gotta say I'm finding it hard to accept their word that they're responsible for Facebook. I haven't paid much attention but I'm under the impression that they primarily just deal in DDoS's and other crude attacks, and I have a hard time imagining that they could cause a large enough DDoS to affect the massive juggernaut that is Facebook. Especially since Facebook just came back up and is now perfectly responsive and showing no signs of being under strain.


    > they could cause a large enough DDoS to affect the
    > massive juggernaut that is Facebook
You don't need to send Facebook more traffic than it normally gets, you simply need to nail a pain point, and cause an unusual traffic event.

There's (I assume) no chance that it's been done using Slow Loris, but that's a good example of doing something unusual to deplete a resource in an unusual way.

I would tend to agree that it's unlikely Facebook have simply been flooded off the internet, but there are many other ways to perform much more targeted DDoS attacks, and presumably Facebook haven't mitigated against _all_ of them.


Good point. Although I'd say that any such attack like that is still unlikely to cause the entire site to become instantly unresponsive (or to serve static error pages as other people have suggested; personally, I just never got the site to load). And similarly, it would not be expected to come back to normal operating behavior a few minutes later. So yes, I'll grant you that you can DDoS even large sites without needing pure brute traffic, but it still doesn't seem particularly likely here.


Yeah, it doesn't seem like a DoS since facebook was reachable and serving error pages, and instagram was reachable and serving blank pages. It _does_ seem like an intrusion or other security incident though because I'd be surprised to learn that instagram shares lots of critical infrastructure with facebook. It seems more likely that someone hit the panic button for both sites.


Some sort of security incidence does seem more plausible than a DDoS, although I can't think of what would affect Facebook, Instagram, Tinder, Hipchat, and AIM simultaneously. I'm also having a hard time imagining what sort of security incident would result in Facebook deliberately shutting down their web presence, even for a few minutes. And all of the other potential attacks, such as DNS or CDN, seems like it a) wouldn't affect everyone simultaneously, and b) wouldn't even work because sites like Facebook don't have a single point of failure, there's always backups and backups for the backups.


Instagram runs within the Facebook data centers. They had some challenging scaling issues within Amazon Web Services and moved their code base to Facebook's infrastructure.


Tinder uses Facebook's API heavily. It's not surprising it went down with Facebook


I wasn't even getting DNS resolution for any of the affected sites.


Yeah, normal access patterns to FB are already pretty much as big as a DDOS on a regular website


For their service infrastructure, certainly. Napkin math suggest that they have to sustain peak floods in the ballpark of 500M concurrent clients. Give or take 100M.

However, if you could cripple DNS propagation/resolution from their systems to the next-hop upstreams, it would show up as services being very slow to respond. They do serve content from multiple domains.

I'd be interested to know how Google's 8.8.8.8 infrastructure defends them from DNS level manipulation...


Um, snowstorm?


Simple Bayesian analysis says it is them as they do not have a record of lying. Why would they? There is nothing to gain and everything to lose.


Nothing to gain? Taking down Facebook seems like something that would bolster their rep significantly. If they think they can claim credit for it and get away with it, then there seems to be little reason why they wouldn't.

I guess we'll just have to wait and see if Facebook makes any public statements as to what the cause of the outage was.


    > seems like something that would bolster their rep
    > significantly
And then irretrievably kill it if they were shown to be lying.


Except that proving they didn't do it turns out to be much harder, because any amount of contrary announcement or evidence can be handwaved away as lies and misdirection. It's roughly akin to trying to talk to conspiracy theorists about 9/11.


How does simple Bayesian analysis suggest that? :-)


It might, but with horrific confidence. Someone just felt like saying Bayesian analysis.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: