New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Do you use IPv6?
This discussion has been closed.
Comments
Sorry, I meant for non-IPv6-compatible clients. Fixed.
They're needed for all the edge nodes... There's a LOT of load balancers, CDN nodes, etc across lots of data centers and PoPs around the world. Every large ISP has Facebook CDN nodes in their data center via the FNA (Facebook Network Appliance) program, but I'm not sure if those use the ISP's IPs or Facebook IPs though.
My VM doesn't have a public IPv4 address associated with it. The host node has a single shared IPv4 which I don't use. Does it still not count?
Those IPs are shared via NAT though. So even if I said to Cam or WebHorizon that I don't need the IPv4, other dozens of customers would still be using them anyway.
I use it when my mobile carrier throws me on it, that's it. It doesn't do anything that IPv4 can't, I receive no benefits from using v6 over v4. There is no part of the internet that I need which cannot be accessed over v4. In it's current state, it's novelty and nothing more. I'm all about wasting time but I have other hobbies I enjoy more.
In my experience, the "resistance" is vastly over-stated, and mostly by a few noise makers. Relatively few who dismiss v6 or rail against it operate what I would consider "real" networks. Sure there are lots of reasons one might feel less than satisfied with IPv6, as you might for any technology, but most recognize that "perfect is the enemy of the good enough" and so, even if grudgingly, accept IPv6 is the path forward and deploy it.
The criticisms arise from two distinct positions. One is technical and typically involves all sorts of complaints about how the design of IPv6 is inherently flawed. This is usually accompanied by various suggestions, which range from an IPv4-forever stance to claims some other alternative would be better, often to include "unique" and "novel" design ideas the resister has. Personally, I tend to dismiss these folks as naive and out of touch with reality, but that's just my opinion.
The other position can be a bit more nefarious and may act like the the former type in disguise. This is relatively small and consists of an what in other industries might constitute a lobby. I'm referring tho those who quite handsomely profit from the scarcity of IPv4 address space and relative high value of IPv4 addresses. If IPv4 were to become largely unnecessary, a pretty sizeable source of revenue for some people goes away.
I suspect this is just an artifact of how IPv6 is being deployed by these providers. Despite IPv6 being effectively a "middle aged" protocol in terms of specifications, deployment expertise is still in it's infancy for many and practices vary widely. I have lots of little systems that are given /64's o more and I have well established providers handing out only a /128. shrug Despite the array of differences in assignment policies, I'm finding that providers that do not provide any IPv6 addressing are clearly in the minority these days. Even in the low-end world.
There is a cost to support IPv6 sure. How much is hard to say. It depends on a lot of factors. We can consider a few. One is simply the allocation costs from the RIRs or wherever your providers get blocks from. It is not terribly expensive, but for some small low-end providers maybe $1000/year is a burden? Another is network and systems support costs. While most hardware and software has sufficient v6 support these days, someone still has to do addressing strategies, configurations, monitoring, and troubleshooting. Even if the cost is only time, that isn't free. Not many people are worrying about IPv6-based DDoS these days, but what costs would that add? Probably plenty if you want to provide symmetrical protection of v6 and v4 blocks.
I think a much more interesting question would be, why do you deploy IPv6? A really compelling answer was given by Comcast a decade ago. If they didn't deploy IPv6, it would have cost them tens of millions of dollars to obtain the necessary IPv4 addresses to do what they do. The hosting business is not Comcast, but in cases where you can start running IPv6 instead of and without IPv4, there may be a real economic case to be made for doing so.
So, at the end of the day facebook is just yet another large IPv6 evangelizer who actually still uses massive pools of IP4 addresses.
And sorry, no, I do not care at all about them using IPv6 inside as they could as well use 10.x.x.x/8
Hehe, not really, but I like you, so I'll give that point
Do you expect users that are on IPv4-only networks to not be able to access Facebook? IPv6 adoption is relatively high in the USA (thanks to all the main telcos having IPv6-only or IPv6-primary networks), but it's very low in the UK and a lot of African countries for example. The IPv4 edge nodes are mainly just for people connecting from IPv4-only networks.
Sounds as if IP4 users were a small minority of exotic weirdos and people living in the bush.
So, if IPv6 usage is so great, facebook could certainly hand back massive amounts of Ip4, no? What's the actual numbers? 80% IPv6 and 20% or even less IP4? (Haha).
Can you kindly provide tangible data about the IP4 fb held say 5 years ago and today?
I prefer IPoAC with IPv9.
Everytime I want to get a fresh linux iso, I send a chicken to the bay.
Not really, apart from that they're RackNerd's backbone.
I didn't mean to imply this anywhere... There's definitely a lot of people with IPv4-only connectivity.
Globally, 33.8% of traffic to Facebook is via IPv6, and it's growing a few percent per year: https://www.facebook.com/ipv6/?tab=ipv6_total_adoption
I'm not actually sure how to do that, sorry
I think that's true to some extent, sure.
We share some opinions.
That's an excellent point.
That's good to know. For me IPv4 is secondary.
Makes sesne.
Ah, the flip side. Interesting take.
Thank you for a detailed and informative response. If there was such a thing as winning the Internet, you'd be the winner today.
TL;DR as I presumed, facebook makes lots of IPv6 evangelizing noise but actually still uses mainly IP4 at the edge (~ towards the users). It would probably interesting to look into the question of how many of the IPv6 users of fb actually had a choice/say -vs- some provider just forcing it upon them.
Maybe we should also make another poll with hosters: How many of your customers really (as opposing to making "advocacy noise") demand IPv6 (as in 'instead of IP4')? Does it reach 2 digit percentages? I have doubts.
There are 13,332 active clients in the database. Their sub-users should be assumed to be exponentially higher. Demand for IPv6 has never exceeded the request of 3 users. If I'm mistaken on that figure, it's still under 5.
One can hope.
On BuyShared I think we have maybe a half dozen users with IPV6, and I'm sure I recognize all of them from these forums.
I'm not going to get on the pro/anti IPV6 wagon, I think BuyVM has done a pretty solid job of supporting it. With that being said, actual real-world usage is incredibly low. It's still in its hobbyist stage.
Francisco
What's with your hard-on for IPv4's held by FB? He already said they're not wasting them and using them in the bare minimum required edge case. You're just wasting his time for no reason.
FB is one of those companies that has a Post-It notes budget higher than the revenue of your company.
And you, yet again, totally missed the point. They NEED IP4, and plenty of them-
If that were true, then what? What were that proving?
We've all seen enough corporate giants who were hacked, and sometimes due to moronic errors.
Why is this even a question or being discussed? Of fucking course they need them. They'd need even more if they weren't using IPv6 primarily.
That they will use and need a lot of IPv4. Seems you agree with me and you were trying to make some point that's irrelevant to this discussion.
What does that have anything to do with IPv6 utilization?
I don't even consider providers that don't offer IPv6. No IPv6 in 2021? I'm not interested in paying you money.
no
Because?
Uhm, having a huge budget == needing a lot of IP4? Exotic "logic" ...
Nothing - before you asserted that having a huge budget == knowing and being right.
So what?
I'm not a provider, but if I happened to be one I'd think "OK, I accept that and wish him, the member of a really small minority, good luck and success to find a provider with IPv6"
... and those would quickly become far less if IP4 proponents pulled off the same attempt of blackmailing just the other way around "If you support IPv6 I won't buy. Bye"
Which is fine. It's not like it's hard to find providers that have a proper dual stack setup.
Dual stack? So you want your beloved IPv6 plus still an IP4? Why so if IPv6 can do everything IP4 can plus some more? Maybe because you want IPv6 as a weird hobby but also IP4 because you want people in the real world actually connect to your server?
Oh well, ...
IPv6 is for poor people that can’t afford an IPv4.
Realistically I suspect that's the reason v6 is more socially popular on LET than anywhere in the real world. Seeing the rising cost of v4 threatens cheap deals.
Yeah, in seriousness, I do think that makes sense. Especially with such large chunks of 4’s already sold in bulk ages ago - growing infrastructures need those IPs to come from somewhere. But if your local infrastructure runs great on IPv4 already, there isn’t much reason to go into IPv6 support - especially as it means introducing risks and new hardware with limited immediate benefit to clients still wrapping their heads around v4’s…
My infrastructure does not run great on IPv4.
I have a socket problem that I couldn't solve with IPv4 NAT, but runs great on IPv6.
This works great on IPv6: when I assign a public IPv6 to each container, each node can open the socket independently, and they are correctly recognized as a single flow.
It doesn't work with IPv4 NAT, because NAT would manipulate the source port number, and then the sockets end up as two separate flows.
Arguably, I could add more and more socket hacks and iptables hacks to make it work under IPv4 NAT.
But isn't it easier to use IPv6, with no NAT causing headache?
If I give a public IPv4 to each container, I can get it work in the same way as IPv6.
In this sense, I have "returned" at least two IPv4 addresses to the global pool by setting up with IPv6.
Note: I know UDP is connectionless. Here "connect" refers to the connect syscall, which sets the remote endpoint of socket.
Under IPv4, it's set to the public IPv4 of the remote host and port 6363, and there's port forwarding to the container's port 6363 via Docker -p flag.
This was expected and how it was planned from the start. IPv4 only, then both, then IPV6 only. No one thought there was going to be a one-time switch at midnight for the entire world. Jesus Christ... you really refuse to educate yourself that much? The ignorance is smelling up the place.
Yes. My ISP has it and it's how I connect to almost every site.