4 ms·

"IPv6 can generate more than 1 trillion addresses while the current IPv4 can 'only' generate 4.3 billion addresses." From wikipedia: "The length of an IPv6 add

"IPv6 can generate more than 1 trillion addresses while the current IPv4 can 'only' generate 4.3 billion addresses." From wikipedia: "The length of an IPv6 address is 128 bits, compared with 32 bits in IPv4.[5] The address space therefore has 2ˆ128 or approximately 3.4×10ˆ38 addresses. This would be about 100 addresses for every atom on the surface of the earth" Technically 3.4×10ˆ38 is greater than 1 trillion, so the article is correct, but still... :) http://en.wikipedia.org/wiki/IPv6 http://en.wikipedia.org/wiki/IPv6

ghshephard 12y ago
One important difference is with IPv4, you have CIDR, so you can utilize a much, much greater percentage of the 2^32 addresses than you can of the 2^128 IPv6 addresses, which, for the most part, has subnetworks that are /64 (you will never see a /70 IPv6 network). I think what's much more interesting about IPv6 is the number of networks with "effectively" unlimited number of hosts that you can have on them. We routinely have IPv6 networks that have 15-20,000 hosts on a single mesh subnet, and there is zero effort required to support that. Trying to design a network that might have as few as handful or as many as 32,000 hosts on IPv4 subnet would be a nightmare I don't even want to consider - even taking into account the impossibility of network assignment, there is the lack of SLAAC - you end up having to rely on things like DHCP, which, while wonderful for small networks, turns out to be a pain in the butt when you have a combination of mobility and millions of nodes.
- clebio 12y ago
  Would be interested to hear about the work you do. I read your profile, but do you have a blog or such where you discuss these sorts of technical problems?
  - ghshephard 12y ago
    I'm mostly an L3 guy who mixes up a little with L2 - but all on the deployment/Ops. The real magic is done by our Routing and MAC engineers who have been coxing the mesh technology along for the last 10 years. What I'd love to hear from, is somebody who has deployed the Cisco CGR at scale with iTron using RPL and the archrock technology.

ghshephard 12y ago

One important difference is with IPv4, you have CIDR, so you can utilize a much, much greater percentage of the 2^32 addresses than you can of the 2^128 IPv6 addresses, which, for the most part, has subnetworks that are /64 (you will never see a /70 IPv6 network). I think what's much more interesting about IPv6 is the number of networks with "effectively" unlimited number of hosts that you can have on them. We routinely have IPv6 networks that have 15-20,000 hosts on a single mesh subnet, and there is zero effort required to support that. Trying to design a network that might have as few as handful or as many as 32,000 hosts on IPv4 subnet would be a nightmare I don't even want to consider - even taking into account the impossibility of network assignment, there is the lack of SLAAC - you end up having to rely on things like DHCP, which, while wonderful for small networks, turns out to be a pain in the butt when you have a combination of mobility and millions of nodes.

clebio 12y ago
Would be interested to hear about the work you do. I read your profile, but do you have a blog or such where you discuss these sorts of technical problems?
- ghshephard 12y ago
  I'm mostly an L3 guy who mixes up a little with L2 - but all on the deployment/Ops. The real magic is done by our Routing and MAC engineers who have been coxing the mesh technology along for the last 10 years. What I'd love to hear from, is somebody who has deployed the Cisco CGR at scale with iTron using RPL and the archrock technology.