A selling point of A&A is the L2TP service that you can use if your primary service goes down. I don't know how easy it would be to provide that sort of functionality if you remove the "session" element of PPPoE and resort to DHCP. Maybe the easiest way to achieve it would be to have the L2TP and FTTP service allocate different IP addresses and relying on the end user to run a routing protocol to use their subnet on whatever connection they decided they wanted it on.

In reply to a post by jpm:

A selling point of A&A is the L2TP service that you can use if your primary service goes down. I don't know how easy it would be to provide that sort of functionality if you remove the "session" element of PPPoE and resort to DHCP. Maybe the easiest way to achieve it would be to have the L2TP and FTTP service allocate different IP addresses and relying on the end user to run a routing protocol to use their subnet on whatever connection they decided they wanted it on.

VPN.

In reply to a post by XGS_Is_On:

Another question related is are you folks going to use PPP indefinitely given a major feature relies on it? Openreach deliver Ethernet, CityFibre deliver Ethernet, at some point BT Wholesale will deliver Ethernet, they wanted to move away from PPP when GEA started, TTW etc will follow. Are you folks going to be having the CPE putting customer data in PPP purely for the LCP echoes to keep CQM running?

Lower power consumption is super important but per subscriber how are the numbers?

That is one of the things I was hoping A&A might have had, the option to use IPoE to avoid needing an over powered pfSense or OPNsense box to get the 1Gig throughput, due to PPP having to be done in software. It's the sort of thing they should be doing to keep the more technical customer happy

In reply to a post by E300:

That is one of the things I was hoping A&A might have had, the option to use IPoE to avoid needing an over powered pfSense or OPNsense box to get the 1Gig throughput, due to PPP having to be done in software. It's the sort of thing they should be doing to keep the more technical customer happy

The more technical customers (and those prepared to pay A&A rates) will also likely use decent hardware and/or software.

PPPoE at 1Gbps is not hard to achieve in software. It's only an 8-byte extra header; similar work to VLAN tagging (4 bytes).

In reply to a post by E300:

That is one of the things I was hoping A&A might have had, the option to use IPoE to avoid needing an over powered pfSense or OPNsense box to get the 1Gig throughput, due to PPP having to be done in software. It's the sort of thing they should be doing to keep the more technical customer happy

Think hardware accelerated PPPoE has been in SoCs for a while now, the weaksauce consumer routers ISPs give need it to hit the throughput target. Shouldn't need too much to get to gigabit throughput through software though unless using the software you described which uses a kernel with bad PPP functionality: single core decapsulation only, no multithreading.

Whether an ISP should be doing something as major as arranging the connectivity and installing BNGs to handle a subset of customers with broken software is a tricky one. PPP at worst doubles the cost of handling a packet but on the x86 kit only really an issue with the software you mentioned. The user always has the option to use different software with the same hardware, for free, rather than expecting the ISP to change their network to fit.

No PPP would break the current implementation of CQM, too, which I imagine is a major issue given how attached these folks clearly are to it.

Giving customers the option will probably have to wait until one of their wholesale suppliers announces removal of support for PPP and they've no choice but to use something else: I think the network, hardware and software is very much engineered around PPP at the moment.

In reply to a post by candlerb:

The more technical customers (and those prepared to pay A&A rates) will also likely use decent hardware and/or software.

PPPoE at 1Gbps is not hard to achieve in software. It's only an 8-byte extra header; similar work to VLAN tagging (4 bytes).

Remember how much work can be offloaded to a half-decent NIC relative to all-software implementations. NICs can take some of the work handling tagged frames off the CPUs.

The software in the distributions mentioned in the post you're replying to handle PPP on a single thread while they'd split handling of TCP/IP and UDP/IP across all cores which is I guess the discrepancy and why PPP requires much meatier hardware on those specific software packages.

The FTTP900 LNS definitely are Cisco. Unchained have some older Firebrick gear, and do do minimal ping based CQM using a Firebrick to my gateway, but I understand they went with Cisco in order to offer fast FTTP services. If there is something in particular you need to know I'd just reach out to them, they're very responsive.

I was also between quota levels on AAISP, so I would spend one month on the high quota and would then eke out the rollover on the low quota for 2 months. It mostly annoyed me it was so inflexible and blunt and I felt dutybound to not pay for more than I needed if it was being monitored... I don't actually need the £20 it was saving me...

No jimbof mate, as Alex (aabloor) said, the Firebrick FB9000 service the FTTP 900Mbit lines - not Cisco.

He’s not talking about AAISP

I was responding to E300's enquiry about Unchained ISP's use of Cisco LNS when they also mention Firebricks. I'm aware AAISIP use the FB9000s (I was on their FTTP900 service for a little over a year).