TaikiSan

5 ms·

1. 1.▲
   by TaikiSan 8y ago

   I'm not sure what you're referring to. The radix tree is stored within the agent context which is running within our user's application. Because of this architecture, making the radix tree larger make the memory footprint of
2. 2.▲
   by TaikiSan 8y ago

   Possibly! We didn't consider databases because this logic is implemented in agents injected in our users' applications. On top of limiting what we had access to, it also put a lot of pressure on us to minimise our CPU & memory
3. 3.▲
   by TaikiSan 8y ago

   We didn't spend too much time on the specific of our choice, because we chose to make something more widely usable (this is how radix trees work) than the specific product evolution we did. That being said, we probably were a bit light
4. 4.▲
   by TaikiSan 8y ago

   Absolutely! This would have been a problem when a IP range was blocked however. You would have to test 32 subnets for any given IPv4 (or 128 for IPv6!). We knew it was not ideal, but it had the least downsides at the time.
5. 5.▲
   Demystifying Radix Trees: How Radix trees made blocking IPs 5000 times faster(blog.sqreen.io)

   179 points by TaikiSan 8y ago|73 comments