It's interesting how seasonal this data is. There's a massive spike on weekend traffic.
Is this because there's a lot less overall traffic on the weekend and the IPv6 part is fairly constant, or is there some other reason why IPv6 traffic should spike* during the weekends?
*As a proportion. It jumps about 50 basis points on the weekend.
I was puzzled about this at first, but my guess is that the weekend spikes are because people are going home and accessing Google's services from residential networks where many ISPs have started to roll out IPv6. Then on Monday people go back to work where there may be less IPv6 access.
That said, I recently had to disable IPv6 on my parent's ISP for bandwidth reasons. They live in France and are served by Free[1], who enabled IPv6 a few years ago. This christmas, I was surprised by their terrible bandwidth, which was worse in the middle of Paris than something in a village in the alps! I ultimitely traced it to the IPv6 option, though I didn't do deeper tracing to find out why.
I suppose it's because many legacy backbone routers don't handle IPv6 at their top capacity, as they have HW to accelerate IPv4 but not IPv6. See for example the question "What is the difference between hardware and software IPv6 acceleration?" in this FAQ at Cisco: [2].
A lot of the problem is probably Free and its policies. The way Free bullied, badgered and French-protectionism'ed its way into the various peering agreements (or lack thereof) could explain why v6 connectivity is worse.
I will say just from a user standpoint, ipv6 is amazing. I recently made the switch to ipv6 on as many of my devices as possible purely because I can then access all of them directly, no network finangling.
Since all my devices have publicly visible ipv6 addresses, I can access them e.g.
because I use tunnel brokers, sometimes the speeds not amazing, but the fact that I can connect at all makes it awesome. Also, those example hostnames are fake : )
Yes, it really is that useful. The ability to directly connect to machines on my home network without mucking with VPNs, port forwarding, etc is fantastic. The key thing it does is reduces friction when creating or using anything that could be accessed from off the local network. The ability to assign static addresses to custom DNS entries makes it even better. The more services that throw up IPv6 support the more useful it gets. If I could pull from Github over IPv6... or run SSL Labs TLS tests against an IPv6 VM...
In particular, it is better than other VPN/port forwarding type solutions because it works transparently and universally. It works on your phone (where VPN can be more tricky). It works in environments where you aren't allowed to install special relay software. It works when you are a guest somewhere else without them having to change their network setup.
As for cloud services... yes and no. In a world with direct addressing you open up many opportunities for innovation in the home cloud appliance market. However, many cloud services are made affordable because they can leverage shared infrastructure, sell usage metadata, or add and remove capacity instantly. Whether home based solutions could be competitive in price and reliability without those abilities. I see some categories of service that it makes sense for, others, less so.
Security is a concern, yes, but one that can and should be addressed by good security tools and practices (firewalls, automatic security updates, good backups, not running services you don't use, etc) not by crippling your network with NAT.
I guess I had become so used to NAT it just seemed the natural order of things. I do think the security is still a major issue but then I like building my servers as if they are alone on the Internet so it makes sense
And yes, a good point about the reliability of cloud services - but still at 50 bucks for 2 TB and iCloud costing me 20 bucks a month I think Shared infrastructure savings have a long way to go !
Large capacity media servers are something I think makes sense as a home appliance. I also think that if people end up willing to put their money where their mouths are that there will be a market for devices that provide cloud like services with the privacy of local hardware.
Re: security. Cloud services aren't doing a lot better. They are repeatedly broken into and have huge password lists stolen, many have governments have direct programs for extracting datas from them, and someone else can call up their customer support reps and practically ask for your password. Many effective security best practices amount to obscurity (change ssh port number). Large central services are very attractive and lucrative to hack.
yes and no. As long as you are able to get a globally routeable ipv4 address for your home at a reasonable price, no. Once you start being subjected to CGN (either or your end or on the remote end) or the price pressure of IPv4 in your region then you will have a reason. Namely cost and performance.
Because of this I would suggest that if you go router shopping that you should choose a model that supports IPv6 rather than not, even if you don't necessary care which you will end up using. Get a model with a good IPv6 firewall that you can tell to block all incoming connections.
Lastly, I think it is dangerous for "casual users" to say things like "I don't care about direct addressing of devices at home". I firmly believe that universal direct addressing will result in new innovation that might change your mind. :)
The TL;DR of "what a tunnel broker is" is not something that I'm totally qualified to explain, but how I approach it is:
A tunnel broker is like VPN/proxy that allows you to connect to the IPv6 web over IPv4.
It adds latency to my connection (because all my IPv6 data goes from me, over IPv4 to the tunnel broker who then routes my data the rest of the way to where it's going), but for me it's worth it.
In terms of security, my machines are run largely the same as any server: locked down, except when for the things I need. It helps that I only do this for *nix devices as well.
I do a similar thing, work ISP provides native IPv6, home ISP doesn't but I'm using a tunnel broker with my router. It has a nice IPv6 firewall built in that lets me manually allow ports through which is nice. It's fantastic having multiple computers at home listening on port 80. :)
> aren't you worried about security of those devices
Assuming he is having a firewall configured on his gateway then it is not any different from security standpoint than being behind a NAT.
For example in my university, all computers have a public IPv4 address, it still does not mean you can connect to them from outside, but if you need to open a port to the world you don't need to deal with all that NAT crap.
What would be the security issue? NAT does not add any security - the firewall does. Just because you have publicly routable addresses doesn't mean they're publicly accessible.
NAT is a good "disallow everything by default" firewall. If my router's firewall had a "disallow all incoming connections by default, it would be great, although I guess I should manage it with iptables, which is machine-specific.
Most ipv6 enabled home routers have firewalls with a "block all incoming connections except for listed exceptions" feature. This effectively gives you the same security as NAT but without the port mangling (can run multiple machines on the same port).
6to4, SixXS, 6RD, and 6in4 are tunnels that are usable by pretty much anyone anywhere right now, as opposed to native IPv6. It helps if your router supports them, of course.
Cloud storage is really nice when your home PC is switched off, or a hard drive crashes, or your computer catches on fire. Other than those sorts of problems, BitTorrent Sync can already do pretty much everything you want over IPv4.
I deploy tinc[1] automatically in my base puppet config to give me direct connection over normal ipv4. It's a private network though so it won't work from a device I don't control as yours probably do.
I'm curious how much of this is due to mobile devices. For as long as I can remember, I have had IPv6 addresses on both Verizon and T-Mobile (though only T-mobile allows IPv6 traffic as of last year: http://www.extremetech.com/mobile/145765-ipv6-makes-mobile-n...)
I think mobile networks do play a good bit of a role here. My iPhone 5s is on Verizon's LTE network and connects over IPv6 to sites. If you look at the latest World IPv6 Launch measurements from Jan 16, you can see that 40% of the traffic that was measured coming from Verizon Wireless' network was coming over IPv6 - http://www.worldipv6launch.org/measurements/
(Those measurements show the % of IPv6 traffic from various operator networks seen by Akamai, Facebook, Google and Yahoo.)
Wonder if the timing has anything to do with companies upgrading networks/equipment over Christmas/New Year holidays? For the US at least, a lot of workers take off significant amounts of time during then so it's often a good time to upgrade or perform maintenance.
Oddly enough we don't, though we added full IPv6 support to our products (which are management tools for hosting providers) years ago, and years before any of our competitors. So, I feel kinda silly for not taking that to the next step and actually enabling IPv6 for our own website. Gonna get on that problem this week.
This is somewhat unrelated to the stats, but is related to IPv6: If you have experienced issues with YouTube videos because your TelCo ISP is messing with the routing, give an IPv6 tunnel a try. I have found that it works great and my last two ISP's (TWC and Comcast) have not yet gone to the point of deliberately slowing down IPv6 or YouTube over IPv6.
It would be wonderful to see a contrast with non-human internet users -- I seem to recall an article recently saying that bots now utilize more port-80/port-443 bandwidth than humans do, and that the trend will only get more skewed from now on.
Sadly, I downgraded my network from IPv5 to IPv4 this weekend, since I couldn't quickly figure out how to get OpenWRT to hand out correct IPv6 addresses. Maybe I'll work on that again this weekend.
Huhhhmmm. Interesting. I already have a linux WLAN gateway between my desktop computer and my router, I guess I could set up the tunnel there and it would be pretty much hassle-free. Thanks for the link.
Yes. Depending on the type of tunnel and the provider you may have to put your Linux gateway into the DMZ. The one type that does not require this is AYIYA (Anything-in-Anything) from SixXS. Here it will be using putting IPv6 into IPv4/UDP packets which traverse your NAT transparently.
In reality it is more complex as the address space is split in two, the bottom 64 bits rarely has any density, only a few devices. End users get a /64 or a /48 (or a few); a VPS will probably get a /64 by default.
Well that is reassuring. For a second there I thought I read that "3% of the address space was used up" but I knew that couldn't be true. I remember learning about IPV4 back in Computer Science class and having to learn subnets and everything. This seems way more confusing!
At that rate it would exceed 100% in four years. Chances are that there will be something like a tipping point well before that, and it obviously won't exceed 100%, but that increase from 2% to 3% is a 50% increase, and promising in terms of future growth if not in current level of adoption.
IPv6 hater here... I'm armed with my asbestos underwear so flame away.
The best advice about IPv6 is to just disable it. IETF designed a terrible protocol that doesn't even address the underlying major issue: routing table size. Instead they just made it much worse.
You won't see me deploying IPV6 for another 20-30 years at least. Hoping it's replaced by something actually usable by then.
I've got native IPv6 at home and we enabled IPv6 on all services at work (hosting web applications and services) and we have zero issues. IPv6 solves a staggering number of architecture problems for us compared to IPv4. For us IPv6 is a net positive.
Marginal Cost. IPv4 addresses are expensive, IPv6 practically free.
Direct addressability. Removes the need for slow, expensive intermediaries like NAT boxes and certain types of proxies by allowing security policies to be applied independently of network addressing.
Simplified network planning. Enough address space to make an address layout based on your actual needs rather than being forced to carve up your limited v4 allocation in weird ways. You no longer have to guess what the right long term size of your subnets are (they are always big enough) or worry about putting unrelated machines on the same subnet because you are out of space for new subnets.
Is this because there's a lot less overall traffic on the weekend and the IPv6 part is fairly constant, or is there some other reason why IPv6 traffic should spike* during the weekends?
*As a proportion. It jumps about 50 basis points on the weekend.