Hacker News new | ask | show | jobs
by maartenh 1291 days ago
Hm, using Apache APISIX for DDoS protection on a single node? That won't really stop a real DDoS. Not much you can do on a single server, if a botnet is saturating the network links to your server(s), without help from your infra provider.

This setup can be used to prevent the backends from being overloaded, which one can probably already do from a single host, and depending on the speed/amount of work by the backends done, not a lot of bandwidth is required to overload most systems that have a limited amount of request processing capacity.

I would argue that this is load management/shedding though, and not DDoS protection.

1 comments

Hasn't DDoS protection always been a small misnomer?

You can't really protect against it, you can only have enough bandwidth to handle everything.

The root of the issue has always been the definition of DDoS.

Network engineers define it as one thing: a massive amount of abnormal traffic generated by a large number of sources (distributed) that you need adequate bandwidth to soak up without impacting normal traffic.

Software engineers define it as another thing: unwanted traffic that causes resource exhaustion and should be blocked. They're almost always thinking of DoS but refer to it as DDoS.

That's why when TFA talks about DDoS, the example immediately attached to it is rate limiting at what to me (network engineering background) seems like an absurdly low limit (1 request every 5 seconds).

“Mitigation” would probably be a better name (and sounds fancier, but maybe not as definitive).