6

I am studying about internet structure, ISPs and their interconnection. However, I have trouble connecting some terms and understanding the greater infrastructure of internet in more detail. I will explain my view below so let me know if I am wrong.

First things first, POPs are places where customers are connected to their provider ISP, to have access to the internet (or at least (technically speaking) the provider ISP's network).

COs are places where the coaxial cables from houses are terminated. In COs there is the DSLAM that splits the signals to telephone and internet traffic. COs pretty much handle the last mile connection.

So, let's assume that A is customer of ISP_A and B is customer of ISP_B and ISP_A and ISP_B are connected via transit ISP_C. If A wants to send a datagram to B (skipping the LAN functionality):

  1. The datagram will be passed from A's router to the modem. The modem will transform it into a signal and transmit it to the telephone line.

  2. Finally, the signal will arrive to the local CO, where the DSLAM will infer that the signal is internet traffic.

  3. So it will transmit the datagram to the POP of the corresponding ISP (probably the same telco).

  4. Then the datagram will be routed to the POP of the transit ISP

  5. It will follow the reverse route to arrive to customer B.

    ...

Have I got that right?

Thank you

Community
  • 1
konstantinosAR
  • 169
  • 6

2 Answers2

5

You're mixing up a couple of things. Also be aware that Internet structures are very different in different countries.

In either case, there's an appropriate modem at the customer end, connecting to a nearby access unit. In the United Kingdom, these are typically in street furniture or telephone exchanges, joined by fibre to larger exchanges. In most countries they are owned by the retail ISP; in the UK they are run by BT Openreach which tunnels the packets to the retail ISP.

  • A CO is a "central office", normally used in US telephony for the telephon exchange.

  • A POP is a point-of-presence, a term used for all kinds of things. During the dial-up era, a retail ISP would rent some space in some town and make a POP: a set of modems linked back to the ISP's centre via a router and a leased line. "We have POPs in London, Bristol and Manchester so you only need to make a local call." The wholesale ISP would have POPs in the internet exchanges, and these would normally be routers, linked by faster leased lines, between the cities of a country and typically also other countries. "We have POPs at LINX and MAE EAST and MAE WEST and such-and-such peering arrangements.".

But you have the basics right

A's equipment gets it to ISP A, which routes it through ISP C to get to ISP B, which routes it to B's equipment. Very often the ISPs interconnect at Internet exchanges -- essentially a fast ethernet switch in a locked room -- and very often ISPs will also peer with each other directly and with large organisations (governments, universities, big companies) over direct links. You can think of the large organisations as essentially being their own ISPs.

The ISPs learn about other by various routing protocols, most notably BGP.

jonathanjo
  • 16,104
  • 2
  • 23
  • 53
  • So, a client's traffic does not need to get routed through his provider ISP's POP, right? If that's the case, where does his traffic go after the Telco Central Office and what is the main purpose of POPs if not providing a "gate" to customers to connect to the ISP's network? – konstantinosAR Nov 14 '18 at 19:14
  • POP is a term related to the economics of long-distance wires, which vary very greatly region to region. The essential thing is that ISPA connects in whatever way it thinks best to get the packets to ISPB, and that might be through ISPC (which might be geographically diverse with POPs) or it might be that ISPA and ISPB are geographically diverse and join in lots of places at their own POPs. Or it might be that ISPA and ISPB are built entirely with DSLAMs and satellite links are there are no POPs at all. There is no uniform practice nor uniform terminology. – jonathanjo Nov 14 '18 at 20:08
2

Welcome to Network Engineering! Yes, you have the basic concept correct. In step 3, ISP_A will determine that the path to the destination address goes through ISP_C, so it will forward the datagram to C. ISPs can connect to each other through common peering points or through private interconnections. It's important to remember that ISP_A doesn't know the full path to the destination, only that ISP_C is the next hop to get there.

Ron Trunk
  • 66,852
  • 5
  • 65
  • 126
  • I think i am confident with the routing protocols. However, don't know if i am right saying that the internet traffic is forwarded from the CO to the POP of the ISP. – konstantinosAR Nov 14 '18 at 18:42
  • 1
    It depends on the particular carrier. Often, the carrier who owns the CO is also the ISP. – Ron Trunk Nov 14 '18 at 18:45