ISP IP address numbers

Is it true to say that every ISP these days needs, and has, at least as many IPv4 addresses available as it has customers?

I believe not. Though those that only issue static IP addresses obviously do.

This is to see what others think. Another poster has insisted they do, and I disagree.

(I hope said person will wait for a few replies to appear ).

Edit - We are assuming they don't use CGNAT.

Edited by RobertoS (Sun 18-Oct-15 17:25:19)

I would say yes, unless they are using CG-NAT or have a reason lots of customers are dropping their internet connections for significant amounts of time.

TalkTalk for example might get away with it as they are (or were) shipping their routers with power saving mode on which drops the connection after a period of inactivity.

But at this stage any ISP with fewer IPv4s than customers is going to have a bad time.

Edited by Adrianuk (Sun 18-Oct-15 17:22:50)

Sorry, I meant if we exclude CGNAT, and have now edited the OP .

Ah, in which case I would be surprised to hear of an ISP that didn't have enough IPv4s for all of their customers.

There will be a large number turning off their routers overnight but a majority of those will have them back on for peak time.

And even taking in to account the router power saving mode scenario in my earlier post, the periods of time these aren't using an IP will be dwindling due to the increase in the number of always-on devices.

I agree, I think that CGNAT aside, every ISP will have at least as many IPv4 addresses as it has customers.

Edited by Oliver341 (Sun 18-Oct-15 17:52:06)

Roberto

I am sure that most of what is written below is 'Teaching grandmother to....'

I believe they do have as many addresses as customers, although you are right in theory that they do not need that many, only enough as customers with turned on modems. Practice is different!

Logic (with example range sizes you wouldn't use) is:- you allocate a range of IP addrsses to a Host ( Say radius server), /24 would have 256 addresses these would all connect to that host, (for customers)
If you were an ISP with only one host you could have 256 customers connected. but your 257th would not be able to connect. so tends to be over provision of addresses.
If you had 2 boxes you would have 512 addresses and you would dynamically allocate across them to keep them balanced ( could be address or bandwidth that decides).
So less wastage however you are always adding new boxes that therefore have only partially used address ranges.

Each host is routed to by the network for that range and then onward routes across the 'access network' to the port on the last access equipment ( DSLAM that will also have it's own IP address pool ).

You cannot easily move ranges between boxes so there tends to be wastage in each host as over allocation tends to occur. IP addresses are still cheaper than buying extra betwork capacity.

The advantage of dynamic allocation is the flexibility in managing capacity as you can automattically balance the nodes at the edge of the core network as customers reconnect without having to move static IP addresses between hosts, therefore adding new capacity is much easier.

If you controlled ALL the equipment, ( as in LLU) you could use a different form of authentication to a radius type functionality and use the IP address range for the actual port but you still get wastage as not all ports will be in use at any point in time. BUT would use far fewer IP adresses in total.

There must be (is) an interesting debate between a few big hosts in the centre of the network hosting millions of addresses and many smaller ones towards the edge with fewer addreses. With Media cache at the edge to save bandwidth in the core I am sure the edges win at the cost of wasted addresses, this will drive IP6 adoption.

I shall have to read that at leisure, as a quick skim shows it covers several aspects I've never thought about. I'm obviously needing to suck a few eggs, but for more complex reasons than the simple "everyone stays connected" argument.

Thanks for that explanation.