Print

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

[bogof]

In case anyone is wondering, £250/pm basically gets you a dedicated port on a switch, as opposed to the shared model in GPON/XGSPON.

I assume they tweak the GPON allocation so that.you always get your 100Mbit up/down no matter what, which probably explains why the uncontended services appear to top out at 100Mbit/s symmetrical.

[Alex Atkin UK]

I assume they tweak the GPON allocation so that.you always get your 100Mbit up/down no matter what, which probably explains why the uncontended services appear to top out at 100Mbit/s symmetrical.

Which is kinda bonkers though given 1000/100 is cheaper for the same upstream and while not guaranteed I can't see there being contention any time soon.

That said, presumably that 100 guarantee is end-to-end, so you'd be guaranteed it even if the backhaul got contended.

[bogof]

Which is kinda bonkers though given 1000/100 is cheaper for the same upstream and while not guaranteed I can't see there being contention any time soon.

That said, presumably that 100 guarantee is end-to-end, so you'd be guaranteed it even if the backhaul got contended.

There's a diagram that seems to show it skipping a whole bunch of backhaul infrastructure.  I've never really understood how Norwich - London can be a 7ms round trip via BTW - that's about 1500km of latency.  Assume this must significantly improve on it.

[XGS_Is_On]

I assume they tweak the GPON allocation so that.you always get your 100Mbit up/down no matter what, which probably explains why the uncontended services appear to top out at 100Mbit/s symmetrical.

No tweaking on the GPON side just that the 220 up product has a prioritised rate above 100 so should never drop below that.

Openreach have fixed OLT profiles, it's a set menu.

On latency the fibre routes are pretty indirect and follow road routes between exchanges and datacentres via rings. You aren't necessarily taking the lowest latency path around the ring as latency isn't considered a factor by most routing protocols. BTW use or at least used OSPF internally which routes based on bandwidth and manually set costs.

There are also a limited number of handover points between ISPs and BT Wholesale depending which product the ISP has purchased so the route is likely even more indirect still if you aren't using BT's own ISP. They use a different product that uses more of BT Wholesale's network.

Getting to those handover points on a broadband product involves going on sub-optimal routes where L2TP tunnels are aggregated for delivery.

What was the access network the 7 ms was seen on? PON is a 1-2 ms round trip to the OLT. VDSL a bit more to the DSLAM.

On BT FTTP from the wilds of Yorkshire my RTT to London was 6-7 ms. Zen 9-10 even on a London gateway, current ISP 9-ish as everything goes via Manchester but the routes are about as good as it gets: national fibre network to Manchester then to London, no stopping off at exchanges only POPs.

The BT service the PPP session stops most of the time at the local exchange and from there it is across the regular BTW network to London, no BRAS / LTS / LNS from the exchange on.

[Alex Atkin UK]

Yeah when me and my friend were both on Digital Region (Origin Broadband) we basically only had the VDSL latency as everything was in Sheffield.  Using the speedtest.net server in the same data centre as the ISP meant 5ms round trip.

In an ideal world we'd all be connected directly to local data centres instead of backhauling across the country, but its just not practical.

Its long been a pet peeve of mine that now to access a server in Sheffield I have to go all the way to London and back, even though physically I could walk to the data centre.

[bogof]

No tweaking on the GPON side just that the 220 up product has a prioritised rate above 100 so should never drop below that.

Openreach have fixed OLT profiles, it's a set menu

Thanks, very informative as always.  So is it fair to assume that a 100/100 product is provisioned at 1000/220 and the prioritisation means it is guaranteed 100/100?  I'm not sure I'd ever seen a list detailing what the guaranteed upstreams are for the products (it certainly doesn't seem to make it into the consumer guarantees!)

On latency the fibre routes are pretty indirect and follow road routes between exchanges and datacentres via rings. You aren't necessarily taking the lowest latency path around the ring as latency isn't considered a factor by most routing protocols. BTW use or at least used OSPF internally which routes based on bandwidth and manually set costs.

There are also a limited number of handover points between ISPs and BT Wholesale depending which product the ISP has purchased so the route is likely even more indirect still if you aren't using BT's own ISP. They use a different product that uses more of BT Wholesale's network.

Getting to those handover points on a broadband product involves going on sub-optimal routes where L2TP tunnels are aggregated for delivery.

What was the access network the 7 ms was seen on? PON is a 1-2 ms round trip to the OLT. VDSL a bit more to the DSLAM.

On BT FTTP from the wilds of Yorkshire my RTT to London was 6-7 ms. Zen 9-10 even on a London gateway, current ISP 9-ish as everything goes via Manchester but the routes are about as good as it gets: national fibre network to Manchester then to London, no stopping off at exchanges only POPs.

The BT service the PPP session stops most of the time at the local exchange and from there it is across the regular BTW network to London, no BRAS / LTS / LNS from the exchange on.

I didn't realise the PON was 1-2 ms to the OLT.

It's 7-8ms total from Norwich to AAISP's gateway in Telehouse London via BTW.  Zen was very similar via BTW (so long as you ended up in London!), but Zen's own backhaul (in spite of the other well-catalog'd issues I had) was notably more direct, clocking in at 4-5ms to London (as low as high 3's observed sometimes).

I didn't realise it was possible that BT's own ISP would have access to different backhaul options, I was under the assumption they're using the same resold BTW service.  Do you know if that also applies to the other BT subsidiaries such as EE and Plusnet?  I don't think I noted much difference when I moved from EE VDSL to Zen (via BTW) FTTP product for latency.

[bogof]

Its long been a pet peeve of mine that now to access a server in Sheffield I have to go all the way to London and back, even though physically I could walk to the data centre.

Haha, I can see my office across the road from my house (it's almost zip-wireable) but likewise I incur two London round trip latencies to the internet from each location, which means the site-to-site VPN I run between has a RTL of 16-17ms.  In practice it doesn't seem to matter at all, as I get 9MB/sec bidirectional for SMB networking between the two sites, pretty impressive as I'm capped to 80/80 at the office end.  I had considered a fixed wireless link between the sites as I have LOS but I think the hassle / cost factor of getting the building to put it up, plus the equipment, outweighs the benefit (it's a 6 storey managed office block).

[dee.jay]

The thing to remember is that the larger internet exchanges are in the bigger cities, I think this a common thing for mostly any ISP - London is the main real site for internet exchange in the UK - that's just how it is.

[XGS_Is_On]

Thanks, very informative as always.  So is it fair to assume that a 100/100 product is provisioned at 1000/220 and the prioritisation means it is guaranteed 100/100?  I'm not sure I'd ever seen a list detailing what the guaranteed upstreams are for the products (it certainly doesn't seem to make it into the consumer guarantees!)

Yep!

See attachment.

I didn't realise the PON was 1-2 ms to the OLT.

10 of 10 packets received   0 % packet loss   Min: 1.346 ms   Avg: 1.491 ms   Max: 1.640 ms

Shared media so have to wait for a time slot to send my data. This is on XGSPON but GPON is about the same.

It's 7-8ms total from Norwich to AAISP's gateway in Telehouse London via BTW.  Zen was very similar via BTW (so long as you ended up in London!), but Zen's own backhaul (in spite of the other well-catalog'd issues I had) was notably more direct, clocking in at 4-5ms to London (as low as high 3's observed sometimes).

Got it. Symptom of the Wholesale product AAISP use and the path your traffic takes to be collected together with the rest of their traffic nationwide. The Zen backhaul isn't going through BT Wholesale L2TP aggregation hops on the way through, with your data getting packed into different tunnels for delivery, more direct.

I didn't realise it was possible that BT's own ISP would have access to different backhaul options, I was under the assumption they're using the same resold BTW service.  Do you know if that also applies to the other BT subsidiaries such as EE and Plusnet?  I don't think I noted much difference when I moved from EE VDSL to Zen (via BTW) FTTP product for latency.

They don't have access to anything others don't. They use https://www.btwholesale.com/products-and-services/data/broadband/broadband-one.html which looks largely like a rebrand of https://www.btwholesale.com/products-and-services/data/broadband/broadband-complete.html I believe. A&A use https://www.btwholesale.com/products-and-services/data/broadband/wholesale-broadband-managed-connect.html

Plusnet definitely use WBMC. EE I think use WBMC.

[XGS_Is_On]

The thing to remember is that the larger internet exchanges are in the bigger cities, I think this a common thing for mostly any ISP - London is the main real site for internet exchange in the UK - that's just how it is.

100%. Networks are built with hubs and spokes. Even when a network seems to be full mesh it just seems that way: still going via hubs. The UK's major two hubs are focused around concentrations of people: London and Manchester.

Virgin Media's network has a bunch of places where traffic aggregates and leaves their network. They can do this as their customers get their IP addresses via DHCP scopes that are bound to physical interfaces locally: it's a routed network straight away. Most of the external connectivity is in London but Manchester, Leeds and other sites have transit.

If you look at the fibre maps of one of the national fibre carriers, there are very few, it gives a good idea of why things are as they are and it's fine. We aren't the size of the USA and a few ms of latency isn't a big deal for most.

[bogof]

Yep!

See attachment.

Really interesting stuff, particularly the trials of the 1800 / 120 service. 
I hadn't realised all the "consumer" grades have the same service level guarantee.  Not that it's likely to bite around here (given the paltry take-up to date), but you might have expected they would give you a little bit more for buying further up the product range.  Odd quirks there, too - it's way cheaper to buy two bottom-of-the-range 330/50 services and basically the same guaranteed level if you can aggregate them, and maybe have the benefit of some element of resilience to some class of issues.

[Weaver]

@Bogof, I believe Kitz wrote about the difference between BT’s own ISP and everyone else a long time ago. It’s like XGS et al said, BT Retail ISP don’t have any special privileges, it’s just that they buy a different product, one that for some reason no-one else seems to want to go for. The difference is in routing flexibility and topology vs control and centralisation. I’ll let Kitz explain it if I can find the page. What about https://kitz.co.uk/adsl/wbc_wbmc.htm ? I need to find some other pages too though.

[bogof]

@Bogof, I believe Kitz wrote about the difference between BT’s own ISP and everyone else a long time ago. It’s like XGS et al said, BT Retail ISP don’t have any special privileges, it’s just that they buy a different product, one that for some reason no-one else seems to want to go for. The difference is in routing flexibility and topology vs control and centralisation. I’ll let Kitz explain it if I can find the page. What about https://kitz.co.uk/adsl/wbc_wbmc.htm ? I need to find some other pages too though.

Thanks, nice reading.
It sounds like the product BT Retail ISP buy is actually something like a "white box" ISP product (given it is sold as needing zero investment).  Almost like BT retail got the ("independent") network arm to build their product for them to just sell...!
I'm sure there must be some other smaller ISPs using it, though I guess competition is probably fierce with the likes of TalkTalk winning on pricing more often.