IPv6 Subnetting – You and your customer

There’s this great debate in the IPv6 world about how to chop up your allocation into assignments for your customers. Typically, most ISPs are being handed a /32, and general guidelines say to allow for a /48 per DSL/leased line/cable customer.

However a lot of people are asking, why not a /64?  Quoted below is the sort of answer you’re likely to receive on NANOG, by one Mark Smith:

There are a variety of scenarios where customers, including residential, will benefit from having multiple subnets. They may wish to separate the wired and wireless segments, to prevent multicast IPTV from degrading wireless performance. They may wish to segregate the children/family PC from the adult PC network or SOHO network, allowing the subnet boundary to be an additional Internet access policy enforcement point. They’ll need separate subnets if they wish to use a different link layer technology, such as LoWPAN. They may wish to setup a separate subnet to act as a DMZ for Internet facing devices, such as a local web server for sharing photos with relatives. Game consoles may be put in a separate subnet to ensure file transfers don’t interfere with game traffic latency, using the subnet ID as a QoS classifier.

This answer is quite simply unrealistic. It’s the answer of a typical geek with no sense of perspective as to what the average consumer wants. It’s the opinion of what Mark Smith the network engineer and geek would want.

In the real world, most consumers of domestic internet services have absolutely no concept of IP addresses let alone subnetting, VLANs, segregation or quality of service. Most domestic networks are a single flat subnet with NAT to a single IP address and no servers that would require port forwarding, and rarely an IPTV system, but those are usually setup to use special triple-play routers configured by the ISP. Most domestic users just want to be able to plug stuff in and have it work.

Now, people will argue that there are more IPv6 addresses than there are atoms in the world. However that argument isn’t as good when you are assigning 1,208,925,819,614,629,500,000,000 IP addresses for just 2 or 3 devices. It’s a grossly inefficient waste no matter what you say. Not to mention that if you’re one of the big cable or DSL providers with millions of customers, it makes much more sense. Each barely used /48 that you throw out contains 256 /64’s.

As such, I personally am inclined to go for the default of a /64 per customer, but allow for a /48 should they need it. There is absolutely no point in issuing a /48 subnet to someone who is never ever going to use it… it’s just laziness, which is what got us into the current situation with IPv4 in the first place.

  1. Let me kindly disagree with your point of you, and explain why “more than a /64” is necessary for end-users. I would say a /48 is good and a /56 a good compromise.

    First, the “waste” of IPv6 is not a bad think.
    IPv6 ranges won’t miss, even if each human beeing on earth as a dozen of /48.

    There is a choice to do between “many subnets for many users” and “many IPv6 for each users”. But the choice of a /48 per customer won’t be a trouble for the whole Internet. Think about the fact that plateforms who actually use hundred of IPv4 (let’s say, a big web hosting provider) could use a /48 for this too. So the total number of /48 should be less that the actual number of IPv4 needed.

    Second, there can be geek in the middle of average customers. People deserve the right to become geeks, and their children deserve this right too.
    I’ve started to discover the network with a very mainstream french ISP. And I have been able to do stuff like hosting my own server, because the connexion was enough “geek-friendly”: I had a public IPv4, it was possible to redirect ports on my connexion. From basic customer I slowly became a geek, and I’m sure that I am not the only one in this case.

    So, let’s assume you provide a /64 per user. In a average family of basic users, one of the kid feels curious about network stuff. He will become to a point where he may need a /48 (let’s say a /56) to go deeper. How come will he get it, if he needs to ask his parents, if they have to pay, or give a reason, if this may need to break and re-do all the existing network of the “geek to be”, who may have set up a foobar::1 IPv6 on a home server and will need to change it because the new /48 doesn’t match on the elder /64?

    Third, assuming that end-users won’t need many /64 is a bet. You and I cann’t guess how IPv6 and the “end to end” principle will be used. Domotic could need it. People may have friendly GUI to set up one /64 for their PC, with firewalling, and another /64 with smart devices (coffee maker, heater..) that would provide a remote control API from smartphones. Yes, one cann’t be sure this will happen, but you cann’t be sure this won’t.

  2. You say that IPv6 ranges won’t be missed even if every human being on earth has a dozen /48’s. Unfortunately that’s my whole point – too many people are thinking right now, and not in 50 years time, or 100 years time. This kind of thinking is what got us into the current mess with IPv4, whereby even some relatively small institutions were given IPv4 /8’s simply because they thought there were more IP addresses than would ever be required. And look where that got us – an extremely painful and decade-long migration between IPv4 and IPv6.

    In addition, my estimates are based on experience. I work for a business-to-business ISP with many thousands of customers. The amount of those businesses that understand subnetting, VLANs or even IP addresses is tiny and near microscopic when it comes to residential customers. It’s great that you’re learning about networking, but you are one in a thousand. Even I can’t envision a realistic requirement today for multiple subnets in my home and I’m about as geeky as they come.

    What becomes of IPv6 when we look 40 years down the line? There’s 7 billion people on earth today, and if each had a dozen /48’s as you say that would be 84 billion /48’s. By 2050 the population could be 11 billion by estimate. That’s 132 billion /48’s, which would put the total usage of earth at around /12. What if start colonizing other planets? Or introducing nano-size computers… thousands of tiny machines on the head of a pin each with it’s own IP address – imagine every pinhead needing a /48 to itself.

    Why not just be conservative now and save ourselves problems down the road instead of making it our children’s problem? One of the best schemes I’ve heard of recently is to assign a /48 to each customer, but only configure a /64 from within that /48. That means that by default each customer has a /64 but can if needed be easily regraded up to a /48. If we have availability problems in future, the unused remainder of the /48’s can be divided up to service other customers.

  3. I’d go with the /56 or /60. 256 /64’s or 16 /64’s respectivly.

    Having only the possibility for 1 /64 feels way too restrictive in the prospect of future technology that may need to split it up.

    /48 on a SOHO user is a tad too much imho aswell.

  4. What I don’t get is why the subnet is fixed, and at /64. This entire conversation would not have been relevant if you, for instance, fixed subnetting at /90 (274,877,906,944 IP’s per subnet!), and then gave everyone a /64. If you’re not going to use more than a handful of IP’s per subnet anyway, subnetting at /64 seems ridiculous.

  5. Madness. I say give everyone /112 and if they want to step up then they pay £5 extra per month per 16 bits up to max /48.

    Nobody will need to pay the extra except those people with electronic hair with individually addressable strands.

    When IPv4 started nobody could imagine why you’d need to conserve address space.

Leave a Comment

NOTE - You can use these HTML tags and attributes:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>