So, for those that don't know, the last of the version 4 IP addresses were distributed in February. ISPs and other large service providers (Google, Microsoft, et al) are hoarding what IP4 blocks they have. My guess is we will see far fewer 8 address (5 usable) IP block allocations from ISPs over the next months and years. I wouldn't even be surprised if ISPs come back saying that they want to reduce the contracted number of IP addresses.
With that in mind, and in my humble opinion, the ISPs created, or at the very least exascerbated, this problem themselves. The way they have chosen to provision IP addresses has turned this into a serious problem.
As an example, they have chosen to provision the IPs at the client-edge devices instead of in a head-end room somewhere. This means that each client-edge device needs a minimum of 3 IP addresses + whatever the customer needs, in blocks of 1, 5, 13, or more. So for a customer that needs 1 static IP address, they now are using up 4. For customers that need at least 2 static IP addresses, they are using up 8.
They use up the extra 3 IP addresses because they need a network IP, a broadcast IP, and a gateway IP for their modem or device that sits at the customer site. Then they give the customer the last IP, for a total of 4. This creates a /30 subnet or in long form, 255.255.255.252.
So then if the customer needs 2 static IP addresses, they do all of the above, taking up 3 IP addresses, plus the customer's two IPs. But the next highest block available is the /29 block, or 255.255.255.248 in long form. So now they are burning up 8 public IP addresses instead of just the 5 they needed.
Most clients only need 1 static IP address, but because of this subnetting and provisioning scheme the ISPs have cooked up, that means 4 or 8 IP addresses get used.
If, on the other hand, the ISPs did the subnetting or provisioning at the head-end room, there could be subnets of 128 or 256 IP addresses. They'd still lose 3 IPs: 1 to the broadcast, 1 to the network, and 1 to the next gateway in the lineup. But that would be 3 IP addresses lost out of 128, leaving 125 usable, or 3 IP addresses out of 256, leaving 253 usable. This would result in a ton less wasted IPs.
The downside to this is that the client-edge devices would no longer have outside, routable IP addresses anymore. It would just be a bridge device between the ISP and the customer's firewall. By the way, this is how AT&T does their bridged DSL modems. The modem has no IP address, but the subnet mask for the customer is /32 or 255.255.255.255 instead of /30 or 255.255.255.252, /29 or 255.255.255.248, or /28 or 255.255.255.240.
Subnetting was originally meant to allow logical parsing of IP blocks within companies and control broadcast traffic. These days companies use internal IP addresses and NAT, and broadcast traffic is controlled by Layer 2 or Layer 3 switches. The time for subnetting things down to a micron are dead. But rather than fix the problem, we're doing what typical Americans do; we throw away the old system and invent a brand new, bigger (but not better), obfuscated system that will allow for more sloth and laziness on the part of the ISPs and network admins. And then we shove it down the rest of the world's throat. IPv6 sucks. IPv4 makes sense, but it's been mismanaged for years.
This problem was avoidable!!
Everyone that knows next to nothing about IPv6 has been infatuated with it for years.
Everyone that has had a chance to play with it thinks it sucks and doesn't want to transition to it. The addresses and address schemas are difficult to follow and impossible to remember.
This whole IP6 vs. IP4 infatuation is a guage I use to tell where someone is technically. If they think IPv6 is awesome and they're excited for the change, I know they're still a little naive. If they're generally not happy about the whole thing and dreading the change, I know they're past the honeymoon stage on IPv6.
I for one will keep customers on internal IPv4 networks until it becomes next to impossible to stay there.
No comments:
Post a Comment