Question about how the Internet works

Forums › New Zealand Broadband › Question about how the Internet works

adam77

46 posts

Geek

#59359 3-Apr-2010 13:34

This may be silly questions but here goes...

What is the difference between connecting to an ISP using a dial-up modem and a DSL modem...

Dial-up (as I understand it):
*Modem at my house makes a call to a modem located at an ISP premises.
*A data link is established between the two modems.
*Data is encoded and sent over a 'normal' voice connection.

DSL (as I understand it):
*Connection is always 'on'.
*Connection is a data link connection (PPP) between my modem and the ISP premises?
*Does the DSLAM act as bridge?
*How does the DSLAM know which ISP to connect to (manually configured, or configured via my modem)?
*How is the ISP end addressed, as a phone number, like with dial-up?

Part 2, so what happens in an unbundled exchange...

1 | 2

637 posts

Ultimate Geek
+1 received by user: 2

Trusted

#314500 3-Apr-2010 22:58

adam77: This may be silly questions but here goes...

What is the difference between connecting to an ISP using a dial-up modem and a DSL modem...

Dial-up (as I understand it):
*Modem at my house makes a call to a modem located at an ISP premises.
*A data link is established between the two modems.
*Data is encoded and sent over a 'normal' voice connection.

DSL (as I understand it):
*Connection is always 'on'.
*Connection is a data link connection (PPP) between my modem and the ISP premises?
*Does the DSLAM act as bridge?
*How does the DSLAM know which ISP to connect to (manually configured, or configured via my modem)?
*How is the ISP end addressed, as a phone number, like with dial-up?

Part 2, so what happens in an unbundled exchange...

There is no simple answer to this question because it depends entirely on the topology and network design of the DSLAM operator - which may be the retailer or it may be a wholesaler.

For dial-up, you are essentially correct. There is a bit more to it in that your modem is dependent on the PSTN which has its own set of switching and signalling requirements (e.g. use of the IN); your modem and the ISP's modem (typically a NAS/RAS device like an Ascend MAX/TNT, UT Starcom TotalControl, Cisco AS5x00, etc) will need to negotiate ("handshake") to establish the physical layer, then PPP will be used to establish the data link layer. Typically PPP authentication with username and password is used. For wholesale scenarios the NAS may be owned by another ISP/telco/wholesale provider which either provides a L3 handoff to the ISP (and most likely uses proxy-RADIUS), or a L2 handoff over L2TP.

For DSL, it varies substantially. You have two main variants: 'classic' TR-58 model networks using PPPoA or PPPoEoA, and TR-101 model networks using IPoE(oA).

The more classical design has multiple relationships:
- User to DSLAM - this is determined by the physical port on the DSLAM and the ATM VPI/VCI.
- DSLAM to BRAS/LAC - this is determined by the ATM VPI/VCI, or if the DSLAM is acting as a PPPoA-to-PPPoE conversion node, by the VLAN ID (typically, double-tagged using 802.1ad).
- LAC to LNS - this can be a statically defined relationship (in the LAC) or dynamic (via RADIUS) based on username credentials or other credentials (DSLAM port or circuit-ID).

At this point your user-configured attributes (such as user@isp.co.nz) might be the trigger to send the traffic to your ISP; or network-configured attributes (such as the DSLAM circuit-id) might be the trigger, or a combination of both.

- User to LNS - this is typically based on PPP and your username and password; but may be based on other credentials such as the DSLAM circuit-id.

The TR-101 design again has multiple relationships but it somewhat simplified since it attempts to treat the network as End-to-End Ethernet:

- User to DSLAM - again, physical port on the DSLAM and ATM VPI/VCI and/or Ethernet VLAN-ID.
- DSLAM to BNG - this is determined by the Ethernet VLAN-ID (S-VID). This could be an aggregated path (N:1) where all users on that DSLAM are carried within a single S-VID, or it could be a unique path (1:1) where each user is carried in a unique C-VID within a common S-VID.
- User to BNG - typically this is authenticated by the circuit-id only, but you can run PPPoE over a TR-101 style network and authenticate via PPP username-and-password.

A DSLAM is somewhere between a switch and a bridge, depending on security requirements (e.g. preventing L2 user-to-user communications) and network topology/design.

If the DSLAM is owned and operated directly by the retailer then much of this is the same, except there is no "user-port-to-ISP" mapping required because they all (in theory) go to one place.

adam77

46 posts

Geek

#314580 4-Apr-2010 09:22

Wow, thanks for the answer!

Couple more questions...

If an exchange is 'cabinetised' (FTTN), the process is more or less the same except it all happens at the cabinet? The ATM network now 'terminates' at the cabinet instead of the exchange/central office?
In a FTTH scenario, does the ATM network get pushed out to the home or remain at the cabinet?

grant_k

3539 posts

Uber Geek
+1 received by user: 124

Trusted

#314590 4-Apr-2010 09:36

adam77: If an exchange is 'cabinetised' (FTTN), the process is more or less the same except it all happens at the cabinet? The ATM network now 'terminates' at the cabinet instead of the exchange/central office?
In a FTTH scenario, does the ATM network get pushed out to the home or remain at the cabinet?

Wherever the Fibre stops, that is where the DSLAM will be, and hence the conversion from electrical to light-driven comms takes place. I'm not an expert on this, but AFAIK, the DSLAMs in the cabinet are just a miniaturised version of what used to be at the exchange, obviously with a lot less ports though, and much more limited backup power resources (small sealed batteries instead of a huge lead-acid battery bank and genset).

With FTTH, some of the early installations are using GPON (Gigabit Passive Optical Network). So they aren't using ATM any more; the conversion from light-driven comms directly to Gigabit Ethernet takes place in a gadget called an ONT (Optical Network Termination). The ONT will typically be installed inside a small cabinet in the subscriber's home, along with a Broadband Router, an ATA to provide phone service, and a backup battery to maintain power for a brief period in the event of an outage.

adam77

46 posts

Geek

#314599 4-Apr-2010 09:51

grant_k:
With FTTH, some of the early installations are using GPON (Gigabit Passive Optical Network). So they aren't using ATM any more; the conversion from light-driven comms directly to Gigabit Ethernet takes place in a gadget called an ONT (Optical Network Termination). The ONT will typically be installed inside a small cabinet in the subscriber's home, along with a Broadband Router, an ATA to provide phone service, and a backup battery to maintain power for a brief period in the event of an outage.

But ATM will take over when the PON terminates at the cabinet/exchange?

In the case of PTP FTTH, where does the ATM network terminate?

cyril7

9073 posts

Uber Geek
+1 received by user: 2499

ID Verified

Trusted

Subscriber

#314606 4-Apr-2010 09:59

I think you will find that ATM is no longer used in either VDSL2 or FTTH services, its VLANs on ethernet, and thats transported nationally via MPLS. As I understand it the newer EUBA service which is provided for VDSL2, so I presume FTTH as well, operates in this manner and can be serviced on ADSL2/2+ connections with the DSLAM performing the VLAN/VPI translation, so ATM is only used between the DSLAM and your modem then back to Ethernet in your modem.

VDSL2 and GPON can support ATM but their native configuration and underlying fabric is Ethernet, I am sure penultamatehop will correct me if I am wrong.

Cyril

adam77

46 posts

Geek

#314609 4-Apr-2010 10:06

Hmm OK, replace 'ATM' with 'MPLS' in my previous questions.

Samsung

Shop now on Samsung phones, tablets, TVs and more (affiliate link).

grant_k

3539 posts

Uber Geek
+1 received by user: 124

Trusted

#314615 4-Apr-2010 10:16

Thanks Cyril7 for mentioning MPLS. I had not heard of this before; rather I thought that ATM was to be replaced by Telecom's much-vaunted NGN. Maybe MPLS is the underlying protocol of the NGN?

Anyway, to learn a bit more about MPLS, I had a look at Wikipedia, and found the following precis of what the benefits of MPLS vs. ATM are:

In particular, MPLS dispenses with the cell-switching and signaling-protocol baggage of ATM. MPLS recognizes that small ATM cells are not needed in the core of modern networks, since modern optical networks (As of 2008[update]) are so fast (at 40 Gbit/s and beyond) that even full-length 1500 byte packets do not incur significant real-time queuing delays (the need to reduce such delays -- e.g., to support voice traffic -- was the motivation for the cell nature of ATM).

There we have the reason for ditching ATM in a nutshell. Basically, the greatly increased speed of modern networks have made it redundant.

cyril7

9073 posts

Uber Geek
+1 received by user: 2499

ID Verified

Trusted

Subscriber

#314620 4-Apr-2010 10:31

Hi, yes as I understand it the NGN is very much only metro lan type setup on MPLS, again penultamatehop is the man with the exact details.

But as I understand it over time all the current UBS services that ISPs use to deliver traffic to the DSLAMs gets moved to the EUBS services, as I understand it most ISPs are already delivering only ethernet to the handover points.

Cyril

adam77

46 posts

Geek

#314622 4-Apr-2010 10:36

Although I agree with the point the paragraph is making, the use of 1500 bytes in that Wikipedia quote doesn't make sense.
It seems to say 'cell switching is no longer needed because 1500 byte cells do not cause delays on modern networks'.
Shouldn't it say 'cell switching is no longer needed because max length 65KB packets do not cause delays on modern networks'?

adam77

46 posts

Geek

#314625 4-Apr-2010 10:44

adam77: Although I agree with the point the paragraph is making, the use of 1500 bytes in that Wikipedia quote doesn't make sense.
It seems to say 'cell switching is no longer needed because 1500 byte cells do not cause delays on modern networks'.
Shouldn't it say 'cell switching is no longer needed because max length 65KB packets do not cause delays on modern networks'?

okay, scratch that, i got confused

cyril7

9073 posts

Uber Geek
+1 received by user: 2499

ID Verified

Trusted

Subscriber

#314626 4-Apr-2010 10:46

I think its trying to say that in older times smaller ATM cells (188byte) were more appropriate to the slower networks of that time, now with faster networks to switch larger ethernet frames is more efficient.

Cyril

Lenovo

Shop now for Lenovo laptops and other devices (affiliate link).

adam77

46 posts

Geek

#314630 4-Apr-2010 10:49

cyril7: I think its trying to say that in older times smaller ATM cells (188byte) were more appropriate to the slower networks of that time, now with faster networks to switch larger ethernet frames is more efficient.

Cyril

yeah i was comparing ATM cells with IP packets instead of Ethernet packets/frames, doh!

grant_k

3539 posts

Uber Geek
+1 received by user: 124

Trusted

#314632 4-Apr-2010 10:54

adam77: Although I agree with the point the paragraph is making, the use of 1500 bytes in that Wikipedia quote doesn't make sense.
It seems to say 'cell switching is no longer needed because 1500 byte cells do not cause delays on modern networks'.
Shouldn't it say 'cell switching is no longer needed because max length 65KB packets do not cause delays on modern networks'?

1500 bytes is the standard used with Ethernet, and has been for many years AFAIK.

To send a single 65kB packet is not practical; it needs to be broken up into smaller packets for transmission, which is handled by the TCP layer with the actual delivery of the packets being handled by the IP layer (if my memory is correct). It's been a while since I read the textbooks on this stuff, but that is my understanding.

PenultimateHop

637 posts

Ultimate Geek
+1 received by user: 2

Trusted

#314763 4-Apr-2010 17:27

adam77: If an exchange is 'cabinetised' (FTTN), the process is more or less the same except it all happens at the cabinet? The ATM network now 'terminates' at the cabinet instead of the exchange/central office?
In a FTTH scenario, does the ATM network get pushed out to the home or remain at the cabinet?

As pointed out by others, the ATM network is pretty much deprecated in favor of Ethernet and IP/MPLS. In the FTTN scenario, the aggregation network (Ethernet or ATM) stops at the cabinet DSLAM. In the FTTH scenario, pretty much the same, except the last mile is now fibre with some fibre technology (GPON, EPON, Ethernet) on top of it.

grant_k: Thanks Cyril7 for mentioning MPLS. I had not heard of this before; rather I thought that ATM was to be replaced by Telecom's much-vaunted NGN. Maybe MPLS is the underlying protocol of the NGN?

Anyway, to learn a bit more about MPLS, I had a look at Wikipedia, and found the following precis of what the benefits of MPLS vs. ATM are:
In particular, MPLS dispenses with the cell-switching and signaling-protocol baggage of ATM. MPLS recognizes that small ATM cells are not needed in the core of modern networks, since modern optical networks (As of 2008[update]) are so fast (at 40 Gbit/s and beyond) that even full-length 1500 byte packets do not incur significant real-time queuing delays (the need to reduce such delays -- e.g., to support voice traffic -- was the motivation for the cell nature of ATM).

There we have the reason for ditching ATM in a nutshell. Basically, the greatly increased speed of modern networks have made it redundant

MPLS (which comes in many forms, but we'll stick with IP/MPLS for now) is the core of most provider networks. 'NGN' is a generic term which means different things to different people (when someone says "NGN" to me, I think "soft switch").

There were many reasons for the introduction of MPLS and replacement of ATM:
1. ATM SARs are expensive, really really expensive, and don't go particularly fast. When ATM became popular in IP networks 155M (STM1/OC3) was fast, and there was no Ethernet alternative. When Packet-over-SONET started offering STM1/OC3, STM4/OC12 (622Mbps), STM16/OC48 (2488Mbps), and STM64/OC192 (9952Mbps) there was little incentive for ATM in terms of bandwidth - most ATM interfaces were a max of STM4 and occasionally STM16. Of course Gigabit and 10Gigabit Ethernet have also altered the game significantly. IP/MPLS is agnostic to the underlying L1/L2 interface.

One of the original motivators behind ATM was to support convergence of voice and data traffic onto a single fabric, and allow for minimal serialisation delay for voice traffic. This is why the small 48-byte cell size was chosen (at significant demand from France) rather than the most IP-friendly 64-byte or 72-byte cell size.

2. ATM has significant overhead in that it uses 53-byte cells, with a 48-byte payload. This means that to carry a 1500B IP packet you need to use 32 cells = 1696B on the wire = 13% inefficient. Depending on the packet size you can see much more inefficient usage (a 100B IP packet needs 3 cells) of the underlying circuit. IP/MPLS doesn't have this level of inefficiency (although it does stack on its own label headers).
3. ATM allowed big IP networks to introduce direct router-to-router circuits through the underlying mesh of ATM which allowed for Traffic Engineering to get more efficient utilisation of the network; additionally you could reduce the number of next-hop lookups required due to more direct adjacencies. MPLS does this as well with Traffic Engineering extensions (MPLS-TE).
4. IP/MPLS has allowed the introduction of numerous new services (L2VPNs, L3VPNs) that weren't possible or cost-effective before.
5. IP/MPLS allowed routers to become faster in terms of switching delay as they could do a lookup on a 20-bit label ID rather than a full 32-bit IPv4 address. This is no longer an issue with ASICs but was an issue on older (mid-90s) generation routing hardware.

grant_k: 1500 bytes is the standard used with Ethernet, and has been for many years AFAIK.

To send a single 65kB packet is not practical; it needs to be broken up into smaller packets for transmission, which is handled by the TCP layer with the actual delivery of the packets being handled by the IP layer (if my memory is correct). It's been a while since I read the textbooks on this stuff, but that is my understanding.

Most 'core' Ethernet networks will be using jumbo-capable Ethernet interfaces, which means they are most likely running at a 9000B MTU or possibly higher. It's worth pointing out that ATM interfaces typically had a 4470B MTU.

PenultimateHop

637 posts

Ultimate Geek
+1 received by user: 2

Trusted

#314766 4-Apr-2010 17:29

cyril7: I think you will find that ATM is no longer used in either VDSL2 or FTTH services, its VLANs on ethernet, and thats transported nationally via MPLS. As I understand it the newer EUBA service which is provided for VDSL2, so I presume FTTH as well, operates in this manner and can be serviced on ADSL2/2+ connections with the DSLAM performing the VLAN/VPI translation, so ATM is only used between the DSLAM and your modem then back to Ethernet in your modem.

VDSL2 and GPON can support ATM but their native configuration and underlying fabric is Ethernet, I am sure penultamatehop will correct me if I am wrong.

Cyril

Yep that's pretty much it - for most ADSL(2) networks ATM pretty much only exists in the last mile, and the DSLAM does provide that ATM<>Ethernet interworking function.

1 | 2