Kitz Forum
Broadband Related => Broadband Technology => Topic started by: Weaver on August 31, 2016, 07:41:01 PM
-
Ten years ago ISPs always used to talk about contention ratios and I assume that at busy times uses wouldn't get anything like the bandwidths that they were paying for, by implication. In fact 7 Mbps / 50 is horrific, and 2 Mbps / 50 is even worse. Stuff of nightmares.
My questions -
* What was the true story back then, in the old days before FTTC availability or even pre-ADSL 2? Were they really overselling that badly? Local aggregate bandwidth into exchanges was that stingy?
* No-one in their right mind is going to pay for 7 Mbps and put up with getting 0.14 Mbps delivered a for the same money? How did they get away with advertising?
And most importantly
* When did things change ? When did this nightmare go away? Assuming it has gone away, please tell me it's not like that anywhere now?
-
50:1 is actually way lower than some isps actually contended at during the horrible adsl max days.
Remember the contention ratio works because alot of the time many connections just sit idle doing nothing.
In the latter years of plusnet under its old ownership they had a contention ratio which if I remember right of over 200:1 which didnt work, they had massive performance problems and it was when they introduced their traffic shaping. I suspect when adsl max got introduced, most isps singificantly increased contention ratio's due to the high price of backhaul, but when 21CN came about they started dropping again, and are now low enough that the major isp's have mostly dropped traffic management.
I would say is 2 prime reasons why ratios are no longer advertised.
1 - its hard to guarantee it given 3rd parties are relied upon.
2 - a good contention ratio is a ever moving target as usage patterns change.
Also size of shared backhaul greatly impacts how much contention ratio is needed.
e.g. if you sell a 10mbit service to 2 people using 10mbit backhaul, you contending at a very low 2:1 but the backhaul is so tiny that it will likely be visible contention. VM suffer from this problem because their node capacities are not much bigger than the products they sell. Whilst DSL exchanges have much better differences e.g. a gigabit backhaul supplying 8mbit adsl connections.
-
So shared backhaul from exchanges has got a lot better since the IPStream Max / ADSL1 / 20CN days?
-
The thing is, some of my local customers used to complain of slowness on the Internet at certain times of day, such as just after the kids came home from school, and sometimes in the evenings. Variable performance, predictable or unpredictable, would drive me mad. I would be happy to pay over the odds for performance guarantees. My wife really needs to get her work done too, and isn't too happy if anything slows her down.
-
yes but bear in mind my example may be better than some live examples, not every exchange has gigabit backhauls. Also that there is still some 20CN exchanges in operation using the higher priced backhaul.
-
My exchange, Broadford, was 20CN until last December. And RevK of AA was indeed bewailing the shocking price of 20CN BT backhaul, discussed in an earlier thread, as you say. Used to charge me double in peak office hours for being 20CN, although they may have stopped doing that now, for some strange reason. In any case I got out of it, went on to 21CN / ADSL2 at the end of last year, the moment I could, and so halved my daytime traffic costs that way.
-
2004-2006 (pre maxdsl) were the worst days for bandwidth.
The demise of 20:1/50:1 contention ratio's started in 2004 when BTw introduced two new forms of charging ISPs, meaning there were 3 ways of paying for bandwidth on the Centrals.
1) Standard based charging - ie with BTw controlling contention 20:1 and 50:1 at the exchange VPs. This was what was advertised as 50:1 etc.
2) CBC - Capacity based charging - contention occurring on the ISP centrals
3) UBC - Usage based charging - ISPs charged for exactly how much bandwidth their customers used.
Plusnet was the first ISP to move over to CBC (http://www.kitz.co.uk/adsl/CBC.htm) which co-coincided with the time they were the only ISP to offer 2Mbps to residential users. They got swamped with 2Mbps accounts, some of them at the time classed as heavy users. Cost of the centrals spiralled and at one point they were having to light new centrals every few weeks, hence why they introduced what became known as the 'Bad Boy Pipes'. Capacity on the centrals was at that time a big issue and in an attempt to keep things under control they were the first ISP to bring in Ellacoya's to help manage bandwidth by using traffic shaping.
Prior to CBC there were some horrendous issues of exchange contention. CBC just shifted the onus on the ISP to control bandwidth at the centrals.
By 2006 all ISPs were using CBC (http://www.kitz.co.uk/adsl/CBC.htm). 20:1 & 50:1 was dropped because by this point none of the SPs were using the old Standard Based charging. No ISP that Im aware of ever used UBC.
ADSL was now available at most exchanges. CBC lowered the cost of more affordable to the masses and prices dropped quite a lot as the newer lighter users helped balance out costs in some respects, but adsl max meant higher speeds and peak time congestion for many ISPs.
Things started to get better after the roll out of 21CN and Market Based charging (http://www.kitz.co.uk/adsl/broadband_access_market.htm).
- 21CN backhaul is more efficient with much larger pipes (20CN MiSP originally used 155Mb pipes, but most exchanges were upgraded to 622's for dslmax )
- OFCOM allowed BTw to reduce pricing on the Market 3 exchanges which helped reduced ISP costs on certain exchanges.
-
anyone remember entanet's STM?
adsl max by the far the dark days of uk broadband.
I never ever noticed congestion on my 1/2 mbit freeserve connection, prior to adsl max the cost of backhaul was a fraction of the cost because the price premium was put on the high speed adsl ports instead which was why 2mbit adsl was circa 80gbp month. But when ofcom forced BT to remove price differentials on burst speed, BT had to be compensated and they were by been allowed to ramp up cost of bandwidth on BT centrals. Everyone moved to CBC because they would not be able to compete otherwise, £80 for 2mbit or £20 for 8mbit? As kitz said this of course moved a lot of the choke points from exchanges to isp's centrals.
Entanet had colour graded their pipes, which were a easy way for customers to see utilisation levels, and someone even developed a app that ran in the system tray to show what pipe you on and the colour :D Central hopping was rife on entanet. But at least they used a fair system which did not pick on certain protocols. Generally for 2-3 hours each evening everyone was capped to 2mbit/sec aside from usually 1 or 2 centrals would have unbalanced lower load so one could hop to have full speed all night.
haha
http://forums.thinkbroadband.com/entanet/3125675-need-more-segs-firing-up.html?fpart=all&vc=1
the old days.
-
I never ever noticed congestion on my 1/2 mbit freeserve connection,
You were lucky. We frequently got it bad at this exchange (155Mb backhaul). Speeds were worse than dial-up at times. This (http://web.archive.org/web/20030606145636/http://www.broadband4blackpool.org.uk/) was my crash course on how BTw operated and how I first found out so much info about their backhauls etc
Then, last Saturday, complaints began to surface and now it seems that everyone we're aware of on the standard 500 home service is experiencing wildly fluctuating speeds. At times the speeds can be worse than 56k dial-up, and at other times they're fine.
//snip//
the issue has been "escalated as high as it can go"
//snip//
The forum staff were contacted on Friday by one of BT's Managers who have recognised that there was a serious problem of congestion on 2 of their Virtual Paths.
BT hoped to overcome this problem by increasing the bandwidth on these 2 paths by 50% ie from 4MB to 6MB. They hoped that users would see an improvement in their speeds.
It became a regular event.. same with 1Mbps...
...and by the time I got 2Mbps - See below for a plot of daily speed tests from my 2Mbps account showing each day at 4pm speeds dropping to 500k, then 300 kbps by 6pm. I was so fed up by this point I went 20:1.
That was the reason one of the PN usergroup members wrote the exchange checker, so people could easily check VP status.
Yep I remember Enta's problems. I think by that time, there was hardly an ISP that wasn't affected. Most ISPs had caps or some sort of throttling or other.
Adslguide was a round of people moving from one ISP to another chasing an ISP that didnt have issues, then a few months later move on when they became congested. :no: Im sure there was only Zen which didnt, but they had caps in place and cost deterred some. AAISP also escaped due to their peak caps and higher pricing.
I note in that thread Jason says
The listing of specific contention ratios were dropped when BT dropped them in Oct 2005. ADSL remains a contended service however.
.
That must have been when BTw started to withdraw standard based charging, because theyd SP's had gone to CBC. I couldnt remember the exact date - only that it had gone by 2006, so I wasnt far off.
-
AAISP did escape but when I was their customer I had a 6 gig peak time cap per month, for almost £40 a month. The usage limit was really low.
Entanet usually had 1 pipe which did not have all segments lit (so smaller than 622), their balancing tended to favour that smaller pipe, and was almost always green with 8mbit speeds. (6mbit on my line). Also I remember people on their business packages would have 2 extra hours in the evening been immune to STM before been affected :)
-
@kitz - Wow. Bad news.
I never had any problems back then. In 2004, I was a Demon (HomeOffice) user, can't remember what the advertised contention was. Just one static IP I think. I'm wondering why I escaped, didn't know quite how bad things were out there in civilisation.
There probably were still relatively few DSL users around here even in 2006. [Back then I spent lots of my time upgrading local customers to ADSL, helping them put in the orders, and installing kit. (All on condition that they agreed to use one of my preferred ISPs, so no nasty surprises and no comeback on me.)] Only a couple of exchanges had DSL, a few business customers were on ISDN.
So perhaps that's why I managed to get away with no slow-downs. In 2006, I got upgraded - to my surprise - to IPStream Max and sync rate shot up from the fixed ~500 kbps to > 1.4 Mbps downstream. Later on, I would eventually go up to where I almost but not quite managed 2 Mbps, so I'd be on 1.75 Mbps d/s.
Then at some point I switched to one of Demon’s business-user deals, which would probably have been 20:1, iirc. [Came with a nice block of static IPs, which was much better.] I had BTW IPStream Max Premium traffic prioritisation, resold by Demon. Can't remember when I first got it. Maybe that kept me safe from the misery of wilt.
-
I don't recall having any issues with either Tiscali or Bulldog back in antiquity.
I'm pretty sure if the data cap were low I'd have breached it and would have never forgotten about it due to the subsequent beating :lol:
Back then we had an old USB Tiscali modem and I used to have to run a 10m phone extension cable out the window if I wanted to use the internet in my room. Come to think of it I only had a laptop back then.... If only I knew routers existed :o the folly of youth.
-
> I used to have to run a 10m phone extension cable out the window if I wanted to use the internet in my room
??? ;D
-
I seem to remember that way back then some ISPs did advertise 1:1 contention service, for a lot of money. (And 5:1 etc?) Am I correct?
-
Anyway, at some point things got better?
The equivalent modern-day pipes must have got a whole lot fatter than the 4 Mbps and 10 Mbps pipes shown in Kitz's 20CN diagram at
http://www.kitz.co.uk/adsl/equip.htm
Are today’s FTTP, FTTC, and ADSL users still seeing time-of-day related congestion?
-
I don't recall having any issues with either Tiscali or Bulldog back in antiquity.
Both were known to have major issues - worse than IPStream providers.
Bulldog in particular was very badly hit because they were selling 2Mbps using 2Mbps datastream VPs (http://www.kitz.co.uk/adsl/adsltype.htm).
The Bulldog forums at one time were in chaos after they started selling cheap datastream. :(
The equivalent modern-day pipes must have got a whole lot fatter than the 4 Mbps and 10 Mbps pipes shown in Kitz's 20CN diagram at
http://www.kitz.co.uk/adsl/equip.htm
Yep that was 20CN. MiSP (http://www.kitz.co.uk/adsl/equip2.htm#backhaul) had/has ATM backhauls which meant 155/622Mbps. :/
21CN is here (http://www.kitz.co.uk/adsl/21cn_network.htm). Uses SDH (http://www.kitz.co.uk/adsl/21cn_network.htm#SDH) on the backhauls so 10/100 Gbps pipes. WDM (http://www.kitz.co.uk/adsl/21cn_network.htm#WDM) will also be in use at the FTTx head-end exchanges which means greater capacity per fibre cable.
Are today’s FTTP, FTTC, and ADSL users still seeing time-of-day related congestion?
You do occasionally still see hot S-VLANs, but its easier for BTw to spot and usually much quicker to resolve than with the old ATM (20CN) VPs.
-
I think bulldog LLU was fine, but their datastream was a very different story.
Those days was the heyday for LLU providers as they bypassed the high BTw costs.
-
Yes their datastream was very bad.
LLU didnt come along until a few years later and coincided with the introduction of OFCOMs Broadband Access Market Classification.
OFCOM ensured that BTw had their prices fixed and BT couldnt reduce pricing until the exchange was de-regulated. This ensured that LLU could undercut BTw by being able to offer cheaper broadband and force competition into the market.
-
I made some comments in the past about sky's capacity. This video its explained and given an idea just how much spare capacity they have e.g. their 8 primary routers individually could handle peak traffic load alone so if e.g. 7 failed there still would be no congestion. Likewise the core network if half of it went down, there still would be no congestion.
https://youtu.be/uQUoO4Wb7s4?t=6m50s
The weakness is the daisy chaining of exchanges, as ignition pointed out on ispreview, those not on exchanges directly linked to POPs (thankfully not me) are at risk of outages if any part of the daisy chaining links fail.
-
I suspect people would be very surprised if they knew how small exchange backhauls were in the ADSL days, and even now how small SVLANs are.
I am not aware of any exchanges that had VPs of 155Mb. 34Mb or less was normal and usually okay.
Even now a 1Gb SVLAN is a rarity and 100Gb doesn't exist as exchange backhaul. Many are on a few gigabit Ethernet links port channelled or load balanced via OSPF at most.
One case that comes to mind is where a 300Mb AAISP customer couldn't get above 200-220Mb. His FTTP had more bandwidth than the SVLAN that backhauls it.
Average usage is still only a megabit or two per connection at peak times.
-
e.g. their 8 primary routers individually could handle peak traffic load alone so if e.g. 7 failed there still would be no congestion.
Are you basing that on the presentation's statement that "Biggest P-router capacity ~= Total Traffic Served"?
Tim does say that, through dimensioning, that it just so happens that the capacity of a single P-router matches the total traffic being served.
However, Tim doesn't go on to claim that the network would still work congestion-free if 7 of the 8 P-routers failed. Instead, he only claims that the network would still handle "peak traffic" (evening) congestion-free if they lost one core site (ie 2 P-routers, one on each ring) or a major fibre route (which I assume to mean a route between 2 core sites, affecting both rings).
Given the architecture (dual-parented PoPs), I reckon that a loss of 7 of the 8 P-routers would immediately lose /all/ service to half of the exchanges, possibly three-quarters, depending on internal details. Being "congestion-free" would be the least of the problems then...
-
I didnt say it wouldnt go down, I just said peak time load was under 1/8 of the capacity of the p-router.
-
I don't recall having any issues with either Tiscali or Bulldog back in antiquity.
I'm pretty sure if the data cap were low I'd have breached it and would have never forgotten about it due to the subsequent beating :lol:
Back then we had an old USB Tiscali modem and I used to have to run a 10m phone extension cable out the window if I wanted to use the internet in my room. Come to think of it I only had a laptop back then.... If only I knew routers existed :o the folly of youth.
Tiscali Shudder Shudder,I joined them back in 2006 iirc, on their upto 2mbps product, and that was never throttled,even P2P torrents, then they re graded everyone to their 8mbps MAX product, that was heavily contended almost 12hrs every weekday ,and virtually all day at weekends, throughput was reduced to sub 1mbps levels dial up was faster, and that usb modem speed touch ST330 iirc was terrible, as was their CS and Tech support , and the 0871 number you had to use to call them, needless to say i started researching my next ISP Ukonline a switch i did not regret,
-
I didnt say it wouldnt go down, I just said peak time load was under 1/8 of the capacity of the p-router.
You weren't saying that either ... and I'd have left that simple statement unchallenged. (*)
Instead, you were making pronouncements about how congestion-free the network would be in the event of certain (huge) failure modes ... which aren't claims that Sky themselves make. Sky's claims are much more limited.
Capacity of one router is meaningless if it isn't fully populated to make use of the capacity. Nor if the core fibre routes can't handle that volume. Nor if there isn't a working router at the other end to receive that data and send it somewhere.
Network capacity is oh so different from router capacity. Network congestion is different from router congestion.
(*) - Perhaps even the simple statement is unlikely to still be true.
When Sky installed the new routers, the fact that the router capacity matched total traffic was happenstance: the hardware was new, with plenty of room for growth. As volumes grow (40%CAGR or more?), that happenstance disappears ... And was always likely to be rendered an obsolete observation.
Now, 18 months on, total traffic may well have doubled (judging from the estimates in the graphs earlier in the presentation). The observation likely no longer applies. It was a red herring, and never useful.
-
wombat if you want to twist my words into something that suits your agenda then go ahead, I know what I said and meant.
Traffic has more than doubled since he was talking about that p-router capacity (early on in the video before he talked about what they did since then which has included multiple upgrades as well as switching to one vendor only). But likewise the router itself has been upgraded with newer kit.
I see you took a dig at VM in another post today and at sky here, no digs at BT of course. Where is the BT presentation for their network? or is it commercially sensitive as usual?
-
I am not aware of any exchanges that had VPs of 155Mb. 34Mb or less was normal and usually okay.
Nope the VPs were much smaller. 155/622 was the size of the backhaul pipe which would be split into VPs of varying amounts of bandwidth. Back in 2003ish a typical VP for a newly enabled exchange was just 4Mbps. Technically speaking they could put 400 x 512kb lines on a 4Mbps VP @ 50:1 or 160 x 512kb lines on a 20:1 VP.
RE BT v Sky network. You cant compare.. they are different types of topology.
Sky's is based mostly on a ring, whilst BT's is meshed.
Meshed allows more resiliency. The info is out there about BTw's network and available to SP's via their login details. You used to be able to bypass the login for some info if youre prepared to do some digging around which is how I got some of the stuff for the main site. Now its all pretty much locked down unless you have a login. Anything from the Tier 1 exchanges is quite well meshed. The CORE PoPs have a lot of resiliency, I don't think Ive ever heard of there being a problem with the actual Core. BT's point of failure appears to be Telehouse ;)
-
Yeah I agree with kitz, it would be stupid if BT were employing 4mbit pipes as physical backhaul, it was software limited, why did they cap it? The regulator made sure they couldnt give themselves free bandwidth to have an advantage over LLU operators. Its all about protecting LLU with ofcom.