Hacker News new | past | comments | ask | show | jobs | submit login

> I have yet to meet an infrastructure engineer or even hear of one who switched to IPv6 on purpose because they thought it was better.

Around 10 years ago the company I worked for used ipv6 to successfully deal with huge number of ephemeral VMs. It was a great solution.

I saw Fly.io using similar approach for their internal networking so I assume they did it bc they thought it was better as well.

So now you have.




So you fall in the more than 65,000 servers camp then.

Of course it's great for ephemeral stuff. It's just that most people don't have that problem and it's a huge pain to use it for anything else.

So no, I haven't.


Spoken like a person who clearly never had to peer two vpcs that just happened to have intersecting subnets

> Of course it's great for ephemeral stuff. It's just that most people don't have that problem

A lot of people run containers these days. It’s not unusual to have endpoints in the tens of thousands or more and there are other considerations that make ipv4 management hard/impossible


Yeah, I have. Not that big of a deal. Most of it's just good network planning. We were able to get around the problem.

And I manage a kubernetes container stack. It's simple, you NAT k8s connections to the outside network, or more likely use an nginx ingress.


> I worked for used ipv6 to successfully deal with huge number of ephemeral VMs.

I find it hard to believe that "huge" exceeds 65,536, which is how many addresses you get with a /16.


Ha, you better believe it - even /8 would be tight for them. Ultimately, it comes down to segmentation creating a ton of waste and you need segmentation to manage ACLs and things like that

With ipv6 you can just assign every actor a /96 and they get 4 billion IPs to play with




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

Search: