Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Weaver]

I'd like to improve my very limited understanding of PPP in Internet access. Is it correct to say that PPP from CPE terminates at the BRAS, but the BRAS then makes a second, relay PPP connection to the ISP’s servers? I may be very wrong. 

By this I mean that the BRAS generates replies to PPP frames coming from CPE in certain cases, but also passes some frames through onwards to the ISP. I imagine that PPP from the ISP heading towards the CPE is forwarded by the BRAS.

Things I believe, based on, errm, what

1. I note that in my case, PPP LCP ‘pings’ are generated by the CQM servers at Andrews and Arnold, and my router replies to them, so this would be either the BRAS-forwarding (or BRAS-transparent) case.

2. I think IPv6 is sent to me inside PPP frames. Starting at the ISP, would this be IPv6/PPP/L2TP/xxx ? (whatever the xxx is? IP?)

Any help, much appreciated.

[Weaver]

Apologies if this is an old topic, if been covered elsewhere already.

[Weaver]

A few extremely vague hints at what might be going in are seen in the page
    http://support.aa.net.uk/Line_Options
for example, in the section below ( italics, my emphasis added ) -
--
LCPFix

Re-negotiates PPP after acquiring the connection from BT, who may provide 1500 MTU when it should be 1492 MTU.

More Info: During the PPP connection when your router initially syncs up and logs in to us, the PPP connection is passed via BT. Sometimes BT can change the MTU. With this option we will accept the PPP connection from BT, but will then re-negotiate the PPP connection with your router allowing the MTU to be reduced. On TalkTalk connections the MTU is always negotiated as 1492 MTU, you will need to select LCPFix and 1500 MTU to fix this, otherwise the LNS will use 1492 MTU when re-negotiating.
--

[Weaver]

That extract seems to imply that BT’s BRAS is a man-in-the-middle at the PPP level, altering or passing on PPP frames between CPE and this ISP.

[niemand]

Hiya.

I reckon http://www.cisco.com/c/en/us/support/docs/dial-access/virtual-private-dialup-network-vpdn/20980-vpdn-20980.html would be useful.

Where people are using BT 21CN and 'old school' BRAS the BRAS functions as LAC and the ISP owns the LNS.

The BRAS is only involved in the conversation at the beginning when it's handling LCP.

You create an LCP session to the BRAS, and using the information there it pre-authenticates you and forwards the LCP session onto the LNS so that you can begin an IPCP session with the ISP LNS.

The BRAS is indeed just routing PPP for you after that. It's functioning as an L2TP router, decapsulating your PPP session from the L2TP session arriving from the MSAN in your exchange, then putting it into a different L2TP tunnel for delivery to the ISP LNS, or to an LTS.

Many exchanges have MSEs now that take the place of the BRAS and can communicate directly with the ISP LNS or can communicate with an LTS on the BT network to reduce the number of L2TP tunnels the ISP's LNS needs to build.

The LTS are a quite different proposition. They function as an LNS to the client and a LAC to the ISP's LNS.

In the basic configuration it's not even man in the middle once IPCP is up between you and your ISP, it's just functioning as a router but rather than forwarding via Ethernet or whatever it's taking traffic arriving in one L2TP tunnel and putting it into another. In the case of the LTS it's very much man in the middle action.

Source: Had the pleasure of building an ISP's LAC/LNS, and have worked on Redbacks and Cisco BRAS 10ks a fair bit from my ntl and Easynet days.

EDIT: Forgot about LTS

EDIT2: In fact Kitz covers LTS in her article on 20CN.

Home Gateway.

The tunnel endpoints on which PPP sessions from the Colossus Network are terminated and where it joins the ISP Central Pipes is known as a home gateway.

The Home Gateway is a cluster of LTS's (L2TP Tunnel Switches). The amount of switches required depends upon the size and type of Central Pipe used by the ISP - for example a 155 Mbps Central terminates on 2 LTS' whilst a 622 Mbps pipe will terminate on 12 LTS'.

I've no doubt 21CN is using LTS extensively too, but you asked about BRAS, and the BRAS just forward stuff, whether to an LTS or an LNS isn't their problem, they build an L2TP tunnel to it and forward PPP to it and that's their job done.

[niemand]

2. I think IPv6 is sent to me inside PPP frames. Starting at the ISP, would this be IPv6/PPP/L2TP/xxx ? (whatever the xxx is? IP?)

Yep! L2TP uses UDP port 1701 for transport.

The BTWholesale network changes the L2TP tunnels, re-encapsulating the PPP(oE) session, but the PPP session itself should be left alone. After it hits your local MSE, if you have one in the exchange, L2TP is popped off and the stream, now PPPoE, is delivered over a layer 2 Ethernet link to your MSAN/DSLAM where it gets encapsulated in VDSL for delivery to your CPE as PPP inside VLAN tagged Ethernet.

[niemand]

As a slightly different case my ISP is a TalkTalk Business one that is reselling TTB white label.

In the case of someone like AAISP TalkTalk Business use BRAS to deliver customers to them over L2TP.

In my case I connect to a BRAS, too, however it functions as LNS. My connectivity is native IP at that point.

traceroute to www.bbc.co.uk (212.58.246.91), 30 hops max, 60 byte packets
 1  gateway (192.168.0.1)  0.240 ms  0.280 ms  0.292 ms
 2  host-62-24-255-10.as13285.net (62.24.255.10)  5.275 ms  5.311 ms  5.296 ms

62.24.255.10 is I think an Ericsson Smart Edge Router:

Aug  1 01:00:40 pppd[538]: peer from calling number 00:30:88:1E:F3:8C authorized
Aug  1 01:00:40 pppd[538]: local  IP address <Snip, is static>
Aug  1 01:00:40 pppd[538]: remote IP address 62.24.255.10

Company   Ericsson
OUI   00-30-88

Were you on AAISP your PPP call would be handled by the Ericsson above, but it would be doing normal BRAS LAC duties, pre-authenticating you then sending your PPP frames off in their general direction

Apart from it being able to behave like an LNS it does a similar job to the MSEs on the BTW network.

[Weaver]

Wow, for the first time ever, I will soon be able to attempt to draw full protocol diagrams from end to end, with the full protocol stacks at each protocol layer's logical endpoint. Thankyou so much. And the links to the juniper site are brilliant. If I were well I would put all this up onto Wikipedia as an example of one ISP's stacks. The variability is something I know nothing about though and an article would have to have a sufficient number of caveats.

[Weaver]

From Kitz’s diagrams for 20CN, in 20CN the BRAS is a big fat protocol converter, which, for me, amongst many other things, used to covert ATM to L2TP, in my case anyway. If I understand correctly. I didn't know that the link from 21CN MSAN to BRAS was L2TP.

Nowadays I'm on BTW 21CN, since Xmas. Unfortunately there is no TalkTalk Business available at the Broadford exchange, otherwise I would possibly have a choice with AA. (Some AA users choose a mixture: a BTW line and a TTB line, for redundancy.) I'm not sure whether TTB is available anywhere in the Highlands. Don't know why not.

[Weaver]

The traceroutes over both IPv6 and IPv4 to bottomless.aa.net.uk are showing just a single hop from my router, so there’s no detail at all visible.

I think I am now on the Falkirk MSE, 21CN-BRAS-RED3-2401-FK, from info displayed by AA’s clueless.aa.net.uk control panel server, so my using the term BRAS was a little out of date since my changeover to 21CN? Is it acceptable to say that an MSE is a type of 21CN BRAS ?

[niemand]

Pretty sure MSEs aren't Redback kit but Alcatel, and do not use that naming convention, they have MSE in their name.

You may be going direct to that BRAS or it could be aggregating connections from multiple MSEs.

I wouldn't be surprised if Falkirk handles traffic from a bunch of exchanges and terminates them on some BRAS, including connections from its own MSE. The MSEs handle not only broadband, but leased lines and other goodies too.

[aesmith]

Some interesting stuff on Plusnet forums about their "new network".  Apparently where before a traceroute showed Plusnet's gateway as the next hop, on the "new network" there are a number of IP hops (with RFC1918 addressing) visible within the BT cloud before the Plusnet gateway.   That suggests that traffic that was tunnelled before is now natively routed.   

And in true Plusnet style some confusion because one of their tests for connectivity is to suggest the subscriber ping their next hop - but these new intermediate nodes don't respond to ping.

[Weaver]

A nice, concise protocol diagram from the juniper site linked to earlier. Thanks to Ignitionnet for digging the page up

[kitz]

>> That suggests that traffic that was tunnelled before is now natively routed.

By default, Shared WBMC uses L2TP to hide any BT routing including the bRAS hops.

With WBC / Dedicated WBMC the SP has the choice of:-
- PPP termination at the MSE bRAS or
- L2TP to the ISP's LNS for PPP termination.

Plusnet appear to be using
- PPP termination at the MSE bRAS for dynamic IPs, by giving the bRAS specific IP pools to use.
- L2TP handover for static IPs. 

More info in SIN 472.   
Theres some nice 21CN PPP Layer diagrams in there that weaver may find interesting (see section 3.4)