15

As I understand it, public IPv6 addresses allocated by the IANA will have the prefix 2000::/3. These IP addresses will be routable through the Internet.

IPv6 multicast addresses on the other side are prefixed FF00::/8.

My understanding is that the IPv6 multicast addresses won't be routable through the internet. Am I right? If so, is there any way to do one-to-many IP routing over the Internet in IPv6?

kasperd
  • 29,894
  • 16
  • 72
  • 122
Absolom
  • 263
  • 2
  • 7
  • http://en.wikipedia.org/wiki/Multicast_address#IPv6 http://www.iana.org/assignments/ipv6-multicast-addresses/ipv6-multicast-addresses.xml – Zoredache May 29 '12 at 21:28
  • 1
    I might not have explained myself correctly. What I meant was that the ipv6 multicast addresses are not considered like public addresses. So will I be able to access a multicast group over the internet or will it only be supported through private networks like in ipv4? – Absolom May 29 '12 at 21:33

7 Answers7

8

So will I be able to access a multicast group over the internet or will it only be supported through private networks like in ipv4?

I must correct a assumption you seem to have made here.

If all the routers between you and your destination support it, then Multicast can certainly work over the IPv4 Internet. It is simply blocked or not configured in many places. I suspect this is because multicast is not well understood, and many people believe they do not need it. So they simply do not permit it through their firewalls/routers.

IPv6 certainly is capable just like IPv4 of having Multicast work globally. Only time will tell us if people actually permit multicast through their networks.

Zoredache
  • 128,755
  • 40
  • 271
  • 413
  • 2
    Yes but I don't see IANA allocating chunk of multicast addresses like they do for public IP addresses. Doesn't it mean that you cannot reserve public multicast addresses and therefore you wouldn't be able to route multicast addresses over the Internet without the risk of having address clashing? – Absolom May 29 '12 at 21:58
  • 1
    @Absolom, the second link I added in my comment included a list of the permanent IANA IPv6 assignments. It can get a dynamic address using a protocol like DHCP (http://tools.ietf.org/html/rfc2730) or apparently it can just get an address by picking a random number. http://tools.ietf.org/html/rfc3307#section-4.3.2 all 3 options are basically the same as what was available with IPv4. – Zoredache May 29 '12 at 23:52
  • 3
    On IPv4 networks connected to the mbone global multicast works. Mbone consists mostly of university networks. Making available interesting video services advertised via SAP or for example NTP. There are security issues: Clients for ghost installs have shown up from 8 timezones away until work blocked that specific multicast group at the border. – Koos van den Hout Jun 02 '12 at 14:59
  • 5
    I think the major issue providers have with global multicast is that it requires a bunch of state inside the network, so it just doesn't scale up to Internet size. Also, figuring out who/how to bill for the traffic amplification is probably tricky in a large service provider network. – rmalayter Jun 04 '12 at 21:10
5

IPv6 public -unicast- addresses have the prefix 2000::/3 (so far). The allocation for multicast includes accommodation for link-local, variously local scoped and global addressing (as per RFC3307). It's the same basic idea as IPv4 multicast, where a chunk of the 224/4 space is set aside for GLOP addresses and such.

Check out RFC3306 as well, I think it probably more directly answers your question.

This specification defines an extension to the multicast addressing architecture of the IP Version 6 protocol. The extension presented in this document allows for unicast-prefix-based allocation of multicast addresses. By delegating multicast addresses at the same time as unicast prefixes, network operators will be able to identify their multicast addresses without needing to run an inter-domain allocation protocol.

So the idea is that if you have a globally routed /64 that it can be included in the overall group ID to give you something that can be routed globally. In other words, if you already have a v6 prefix and a suitably enabled carrier then you're already set.

rnxrx
  • 8,103
  • 3
  • 20
  • 30
  • But will the ipv6 multicast addresses will be routable over the internet? Say, for example, you want to stream a live video feed over the internet to a ipv6 multicast group. Will this be supported? Or the ipv6 multicast will only work in private networks? – Absolom May 29 '12 at 21:29
  • 1
    Yes - some ranges of the FF00::/8 will be routable globally, others will have limitations. One of the fields defines administrative scope, which gives routers a clue about how far to forward a given multicast. I'll add an example to my answer. – rnxrx May 29 '12 at 21:37
3

The standards allow global multicast routing, but AFAIK currently most ISPs restrict multicast only for their use (IPTV etc.)

mulaz
  • 10,472
  • 1
  • 30
  • 37
  • Could you please direct me to the standard that you are referring to? – Absolom May 29 '12 at 21:49
  • 3
    http://tools.ietf.org/html/rfc4291#section-2.7 ffxe::/16 are multicast addresses with global scope, so reachable globally. – mulaz May 29 '12 at 22:10
2

My own guess is that many ISP's that happen to be cable providers who deploy IPv6 will block it at their border.

It may be cynical of me to think this, but they have a vested interest in keeping the costs HIGH for their content competitors. v6 multicast would allow HBO or any other streaming video provider to transmit one v6 stream to the Ipv6 world over a multicast channel and dramatically cut costs.

Danathar
  • 21
  • 1
2

I see multiple answers assuming you have mistaken what multicast is. You have made no mistake and your question is clear. I asked this question to myself:

Can I multicast across the IPv6 Internet?

Traditionally, such as in IPv4 I would need to request a permanent global multicast address (or subnet) and assign these to my network. This is still possible with IPv6. However, IPv6 is inherently multicast-happy so it would make sense to include some mechanism for me to multicast to you without requesting unique addresses.

The advantages become obvious when you have three people each on two physically separate networks (six people) playing the same network game. The options are to unicast packets to every player (five packets sent for every update), or to multicast (one or two packets for every update): the first packet would be sent to a link-local multicast address for the players on the local LAN, and another packet would be sent to the global multicast address, which the routers would understand are intended for the players on the other LAN. It may even be the case that the packet is sent once to the global multicast address, and the router (or the local clients) know how to deal with that. The latter would certainly be more efficient.

Given how useful multicast is, it would irritate IANA if they had to assign multicast prefixes for everyone who wants to play a network game, or deliver a video conference, or broadcast a live performance to friends, and so on and so on....

The application form from IANA clearly states that you probably don't need to ask for a permanent IPv6 multicast address, which is nice.

Unicast-Prefix-based IPv6 Multicast Addresses

This, of course, has been addressed. The title "Unicast-Prefix-based IPv6 Multicast Addresses" should really say it all: if you have a globally unique IPv6 IP address then you (your computer/device) can assign your(/it)self a globally unique multicast address which is based on your unicast assignment. The requirements are that the software at every point (servers, routers, clients) knows what it's doing. Old routers and lazy ISPs are likely to be the downfall for the next few years.

It was incredibly difficult to find an answer to what seems to be a very simple question, and the closest I could come to find a definitive answer was in RFC3306:

The following are a few examples of the structure of unicast prefix- based multicast addresses.

    -  Global prefixes - A network with a unicast prefix of
       3FFE:FFFF:1::/48 would also have a unicast prefix-based
       multicast prefix of FF3x:0030:3FFE:FFFF:0001::/96 (where 'x'
       is any valid scope).

    -  SSM - All IPv6 SSM multicast addresses will have the format
       FF3x::/96.

Most articles (and answers) on IPv6 multicast focus on local multicast which has pre-defined addresses, and are not very helpful. The clincher is that a client can assign itself a unique multicast address based on its unicast address, and of course scopes still apply:

The scope of the unicast-prefix based multicast address MUST NOT exceed the scope of the unicast prefix embedded in the multicast address.

It's no real surprise that this is difficult to answer as IPv6 connectivity is so rare that testing the Internet for IPv6 multicast ability and reliability is impossible for most end users, and as a result there are not a lot of articles written about it. Indeed, most end users have no idea why they would use IPv6 multicast at home, but the applications are ready and waiting.

This page talks about some confusion brought on by the RFCs, and RFC3956 mentions how certain multicast domains have trouble talking to each other. It may, at this point, be difficult to implement, but I see no reason why the game server (from my above example) cannot assign itself a multicast IPv6 address and inform the clients of this address, and all without having to beg for a static multicast IPv6 assignment.

This is something I would like to follow up myself in the future.

Step 1: ISPs need to enable IPv6. Still.

See also: RFC6308: Overview of the Internet Multicast Addressing Architecture

Community
  • 1
Ken Sharp
  • 194
  • 10
  • How to make it scalable is a challenge. The size of the routing tables in backbone routers is already a challenge. The routing table entries needed for multicast don't aggregate as well as other entries. It may not be feasible to build a CAM large enough to handle an internet where everybody uses multicast. – kasperd Oct 25 '15 at 20:36
  • Indeed that does seem to be the case... initially, but routers that are aware that these are unicast-based should have no more difficulty routing the multicast packet than it does the unicast packet. There is no need for a separate lookup table as the destination is already embedded in the address. All depends on if it has been correctly implemented in the router again. As usual, couldn't find anything on Google which describes the process. – Ken Sharp Oct 25 '15 at 20:55
  • 1
    The unicast prefix which the multicast address was based upon tells you nothing about where the members of that multicast group are. The prefix in a CAM entry can easily cover thousands of end-customers. Each of those end-customers could create many multicast groups based on their assigned prefix. All of those multicast group would lead to a single CAM entry if your approach was being used, but they are not to be routed to the same locations. – kasperd Oct 25 '15 at 21:46
  • It tells you exactly where it comes from. Read the RFC. – Ken Sharp Oct 30 '15 at 06:17
  • 1
    Routers don't need to know where a packet came from. They need to know where to send it. – kasperd Oct 30 '15 at 20:45
  • This is indeed a tricky issue. It's equivalent to establishing a TCP connection to a target, but from that point on omitting the destination address for each packet sent, expecting every router between you and the target to remember where your unlabelled packets were supposed to go. In some ways it's veering away from packet-switching and back towards virtual circuits. – Malvineous Mar 18 '17 at 04:50
  • Routers will have to remember where the packet goes, so we'll probably need more capable routers. But we keep getting more capable routers. It's all hampered by ISPs anyway, like mine. – Ken Sharp Mar 21 '17 at 23:12
0

You confuse two things in my opinion: multicast routing and IPv6 address allocation.

Multicast routing depends on all routers on the path between the multicast source and you forwarding multicast packets to the destinations which want to receive a certain multicast group. In an uncontrolled network with full multicast network any source system can send to any multicast group without any controls, and all systems which have joined that group will receive the data. This has nothing to do with any allocation, just with multicast routing being enabled.

IPv6 address allocation from 2000::/3 decides who can use which unicast address.

So who can use which IPv6 multicast group is for as far as I know only defined in rfc3307.

Multicast sources do not have a lot of control which destinations see their packets, it is quite possible data transmitted via multicast reaches destionations not available via unicast.

Koos van den Hout
  • 1,086
  • 6
  • 9
  • yes I understand the difference but I maybe I didn't ask the question correctly. What I was implying with my question was that IANA is allocating global unicast IPv6 addresses to avoid address clashing. They don't do the same thing for the multicast addresses, so I thought that multicast was not intended to be routed over the Internet. Zoredache mentioned something about generating random multicast addresses to avoid this problem. – Absolom Jun 04 '12 at 17:46
0

I think IPv6 multicast should be routable over the Internet, in contrast to the IPv4 multicast that isn't routable over the Internet.

https://en.wikipedia.org/wiki/Multicast_address#IPv6

ffxe::/16

Global scope

Eligible to be routed over the public internet.

https://en.wikipedia.org/wiki/IPv6#Multicasting

In IPv4 it is very difficult for an organization to get even one globally routable multicast group assignment, and the implementation of inter-domain solutions is arcane.[20] Unicast address assignments by a local Internet registry for IPv6 have at least a 64-bit routing prefix, yielding the smallest subnet size available in IPv6 (also 64 bits). With such an assignment it is possible to embed the unicast address prefix into the IPv6 multicast address format, while still providing a 32-bit block, the least significant bits of the address, or approximately 4.2 billion multicast group identifiers. Thus each user of an IPv6 subnet automatically has available a set of globally routable source-specific multicast groups for multicast applications.[21]

http://ipv6friday.org/blog/2011/12/ipv6-multicast/

Executive summary:

  • IP Multicast was an addon to IPv4, mostly used for delivery of multimedia.
  • In IPv6, Multicast replaces Broadcast and is an integrated and necessary function in your IP network
  • Automatic configuration of IP addresses and DHCP both use IPv6 multicast

There is "Solicited multicast" in the IPv6 for applications. These addresses automatically apeared when on the some interface is created Unicast address. Just use FF02:0:0:0:0:1:FF00::/104, and other 24 bits are the same as in the Unicast address. https://www.rfc-editor.org/rfc/rfc3306

Community
  • 1
Alex
  • 171
  • 4
  • It is doubtful that the ISPs will agree to any single multicast routing protocol or spend the resources necessary to make this happen. Certainly, dense mode multicast would collapse the public Internet, and sparse mode will cost a lot of money – Ron Maupin Jan 05 '18 at 22:01