back to article Researchers seek Internet's choke points

Cable Internet access really is faster than DSL – but paradoxically, cable users get less of the throughput they think they're paying for. That's one of the conclusions of a study* from America's National Institute of Standards and Technology (NIST) which ran the slide-rule over datasets captured under the FCC's Measuring …


This topic is closed for new posts.
  1. LarsG

    So here in the UK we are not alone in the phenomenon of getting less broadband speed than we promise or pay for. 'Up to' is just an advertising fantasy figure.

    1. Danny 14 Silver badge

      ADSL speed is predominantly dictated by distance to the DSLAM - be that FTTC or normal backhaul to the exchange. Cable is predominantly dictated by the local load vs the streetbox to the hub bottleneck.

      so in short, if your DSLAM is overloaded then ADSL will be contended out to lower bandwidth (although you may sync faster or slower depending how the kit is set to throttle). Cable depends on the load both local and backhaul from streetboxes although you should sync much much faster. Have lots of local people but not many streetboxes on the backhaul? You should be ok unless you have the 24hr torrenters on your street. Have lots of streetboxes on a spur? Good luck there.

      The survey seems to be confusing "speed" with bandwidth. The two are not really the same. An ADSL may sync at 4Mb and give you 512k download forever. Cable modem may sync at 20Mb and give you the same 512k forever. Which is better? I'd say the ADSL as it shows it is working at peak efficiency and must be a good connection for the spec. The cable modem in this case is woefully underperforming which shows strain and bottlenecks - due to history of upgrades id say the cable would only get worse...

      1. Roland6 Silver badge

        Re: Line speed/sync rate and efficency

        I don't agree with the logic you used to compare ADSL and Cable to determine which was the more efficient.

        About the only valid conclusion is a Subscriber would be happier with the ADSL connection you describe than the Cable connection, because of their perceptions based on price of the service and expectations of a higher sync rate.

        I'd like to see some research-based comparison of cable network design and it's impact on traffic handling and the service subscribers/end nodes receive verses ADSL, ie. in the same vein as the comparisons between 802.4, 802.3 and 802.5 in the late 80's.

    2. David Hoffman

      Up to speeds.

      As far as the DSL services that the Measuring Broadband America project recorded, most were definitely less than the advertised speed. The issue is if the additional addressing done by DSL, above and beyond TCP/IP, should be counted by the system. If you include the added addressing bytes, then DSL speeds get to be statistically significantly above 90%. AT&T has actually run into a problem with the issue of exactly what, and how, to count data when computing how much of a subscriber's monthly data usage limit is used up each day.

      Initially some very knowledgeable subscribers, using good data transport measurement software, were having AT&T say they used much more data transport than their own measurements showed. AT&T never acknowledged the problem, but did suspend the TOS sanctions for going over the monthly usage limit in several geographic service areas. The conflict is that subscribers only want to measure pure content bytes and DSL providers want to include every byte used to download or upload content. Addressing, checking, and other items take up a significant percentage of a DSL internet connection's usage of bytes during the transfer of content. Should those other bytes be counted in the calculation of how well DSL service providers are doing at transferring data in comparison to their advertised data transfer rates?

      One solution is to do away with advertised speed tiers. You connect the DSL modem, and you get whatever the system can deliver. A provider called Sonic.Net does this. Obviously some subscribers very near to the DSLAM get the maximum that ADSL2+ download can deliver, about 24Mbps. Some subscribers very far away from the DSLAM get dial up speeds. Since there are no monthly usage caps, many subscribers think that is an acceptable compromise.

      Another solution would be for AT&T, and other DSL ISPs, to upgrade their networks, or change network management, just enough to have the Measuring Broadband America project record that they are delivering 100% of advertised speeds. For example many AT&T ADSL2+ modems could deliver significantly above the 6Mbps of the top advertised tier. Unleashing those modems, and the connections to them, would significantly boost the average that the Measuring Broadband America project would record. Similar limit relaxations could be done for the 3Mbps, 1.5Mbps, and 768Kbps connections. These actions would get the average closer to 100%.

  2. John Smith 19 Gold badge

    80% of 5.4mbs 80% of the time across a country of 5 time zones.

    Now remind me, what was that about "superfast" broadband across the whole UK again.

  3. AndrueC Silver badge
    Thumb Up

    Their conclusion is that something about ISPs' network architecture makes cable networks more susceptible to recurrent congestion than DSL networks

    Could it be because a single cable serves multiple properties - entire streets or even groups of streets? At least with DSL the connection is uncontended back to the DSLAM. I'd have hoped a technical study would already be aware of that issue though.

    1. Danny 14 Silver badge

      therein lies the rub. ADSL (as in MAX or 21CN) places dependency on the copper (or aluminium) quality and the choice of ISP as to how they load their DSLAM. FTTC alleviates this somewhat as the mini DSLAM should be up to the task and the fibre backhaul will depend on the ISP.

      Cable might be fine if your streetbox is a "loner" or on a trunk. If it is spur'd to a huge estate covered in boxes then odds are you will fine someone who wants to download 24/7.

      1. AndrueC Silver badge

        I've always assumed that from the DSLAM back there is no difference between cable and DSL. I can't think of any aspect of the backhaul, core or transits that would be specific to either technology. Surely the DSLAM/node takes care of 'decoding' the signal on the twisted pair/coax and both emit as Ethernet.

        Hence - what else could it be but the local cabling differences?

        1. Anonymous Coward
          Anonymous Coward

          "Hence - what else could it be but the local cabling differences?"

          In the case of a UK-style DSL operation, it's how much the network operator is prepared to pay for bandwith between DSLAM and Internerd.

          Yes there's dedicated twisted pair per end user between end user and DSLAM.

          Beyond that there's shared bandwidth between DSLAM and Internet. Sometimes that shared bandwidth might include a BTwholesale piece (sorry about that), other times it may be an LLU operator's bandwidth (which they may rent, in part or in whole, from someone else).

          In either case, more (shared) bandwidth costs more money.

          Local cabling differences are just one factor. Sometimes they'll be the limiting factor, sometimes the limiting factor will be the shared bandwidth.

          1. AndrueC Silver badge
            Thumb Up

            Although the way backhaul is handled in the UK with multiple providers and ISPs all trying to compete on price you'd think that would count against them. Perhaps it doesn't. Perhaps it actually helps. Then again it's hard to tell with only VM to compare against. That has a rep for high levels of jitter and the usage limits imply a lack of capacity but there's also a lot of people very happy with what they get.

        2. Anonymous Coward
          Anonymous Coward

          In my admittedly limited experience

          In my admittedly limited experience, DSL providers tend to be telcos, who are much more used to providing 99.9% or better uptime, fast response if the connection goes down (because the cost of failing to connect a 911/999 call can be literally life or death), better at understanding the concept of provisioning for their customer base, etc.

          Cable companies tend to be more used to 80% uptime, "we'll maybe get somebody out there sometime next week, is that a problem?", and "provisioning? what's that?".

      2. HMB

        I've always had a stunning service out of cable in the UK. At peak time my speeds were always 50Mbps at my last place and I did use the connection quite heavily too.

        All this report really conveys about cable is how US companies have chosen to manage their cable networks. That's not necessarily a criticism against them either. For fast well managed networks you need punters prepared to pay for them.

  4. London 75

    Hmm, dodgy numbers

    Over 80% of speed for 80% of the users is the same as under 80% for 20% of the users. So if cable is more than 20% of the time it may be as insignificant as the following

    DSL slower than 80% for 20% of the users

    Cable slower than 80% for 20.1% of the users

    Very dodgy quote.

    1. Squire James

      Re: Hmm, dodgy numbers

      Better number estimates are further down in the article. Believe it or not, many articles have important information in other paragraphs besides the first.

  5. Anonymous Coward
    Anonymous Coward

    A Damn Solid Link

    We have two leased lines and basic ADSL for web bandwidth.

    I know it's not related but in the last five years we've had about 14 days down time on the horribly expensive leased lines (requiring physical intervention) and about zero on the Plusnet ADSL.

    If someone said to me what is your main trusted line I'm afraid it would be the ADSL.

    1. Danny 14 Silver badge

      Re: A Damn Solid Link

      We used to have an easynet repurposed ADSL 1:1 and I cannot recall it every being down during the working day. The modem was an ancient "console only" thing too that never seemed to need rebooting. We also had a backup ADSL on a different circuit (demon) which also didnt drop out (internet was load balanced as preference to that line) however the download fluctuated on the demon line and never really moved on the easynet line.

  6. sequester
    Thumb Down

    To paraphrase Lewis Black

    If you were a roofer, and you built a roof, and you were 20% off, you'd still be serving time.

  7. johnwerneken

    what a headline

    CHIPS are down? In that case, NO throughput!

  8. Robert Grant Silver badge

    Minimum bandwidth rating?

    Whichever ISP offers this first, I'm going for.

    1. Don Jefe

      Re: Minimum bandwidth rating?

      The local home grown provider offers minimum bandwidth guarantees and they try to stand by them. The problem though is they can't afford to service their infrastructure and their solution is to deem problem area customers 'ineligible for service'. There are only two towns of any size in the county and both were disqualified for service. Comcast came along with their fiber about two years ago and the little ISP is now nearly dead.

  9. John Miles 1

    Different Choke points and ease of upgrade

    For a DSL network you get whatever your sync rate is on the copper line and then the choke point (where subscribers share bandwidth) is the backhaul link from exchange to the core network. This can be eliminated very easily and quickly by migrating to a higher rate backhaul (providing the operator is willing to pay for it). Much the same is true of FTTC.

    For a 'Cable' network the data flows are like a tree fanout across the neighbourhood. The choke point is where they all come together at the 'trunk' of the tree i.e. the head end. You can split it into two trees to remove the 'choking' - but this requires work in the street and lots of re-connections. Difficult, expensive and slow! This is probably the reason why cable networks congest on shared bandwidth, i.e. it takes time and investment to relieve it.

    1. Roland6 Silver badge

      Re: Different Choke points and ease of upgrade

      Both networks are 'trees' in the topology sense, hence both have a head end choke point (although we don't use the term 'head end' to refer to the DSL infrastructure within an exchange). I suspect that much of the difference is down to the strategy used to share out the contended infrastructure and the traffic profiling and buffering strategy used at the head end to manage the internet connection.

      For example, it is relatively simple and efficient to give every cable subscriber a fixed share of the capacity of one channel, thus provide a 'guaranteed' 512kbps download service say, but to give them an on-demand allocation from one or more other channels (with much higher contention ratios) to deliver the higher download rates.

      What is clear, is that we are missing some real comparative network analysis.

  10. Colin Ritchie

    The Following is a true story, only the names have been changed, to protect the guilty.

    About a decade ago I visited a friend on a Nottingham housing estate for a couple of days. He proudly announced that he was now on a cable internet service having given up on a copper 2Mb ADSL line.

    I was interested to see what the difference was and at about 11pm at night it was indeed mucho faster. Unfortunately, during the day and especially between 4pm to 8pm eBay pages would take minutes to load and his torrents would grind to a halt for hours. As for attempting to game online or perform a Widows Update, forget it till well past bedtime for the majority of the estate's kids.

    I presumed that the cable salesman had sold a line to every bloody house in the estate with a PC and the whole damn lot of them were being strained through 1 box at the end of the access road.

    Next time I visited, internet service was now being provided by a 3G ISP direct to his laptop and he'd told the cable company exactly where to stick their bills. It was still slow but at least he was getting exactly what he was expecting and no other bugger was hogging the bandwidth.

  11. itzman

    THis is no surprise.

    IF the choke point is essentially on a per person basis..that is your 4Mpbs ADSL - then when you slam it, it only affects you.

    If however you have high speed connections to the first concentrator - the DSLAM or equiv. - then its easy for a few users to fsck up the backhaul, and everyone slows down.

    FTTC or FTTP is not the fial answer. You need on a busyish network, 1/10th of the backhaul capacity that you have in the customers links. IF you want them to have reasnjable joy. Simply upgrading the last 5 miles doesn't help if you don't upgrade the backhaul in terms of peak speeds. Worse, if you don't throttle you will find that that backhaul is underutilised.

    Or to put it another way, you want to operate at as low a peak to mean ratio as possible. Your customer wants as high a peak to mean as possible.

    And the age old law of engineering raises its head. high peaks = high costs. low means = low income.

    Its why wind power is so expensive, as much as anything, as well

  12. William Boyle

    docis (cable) == shared bandwidth - dsl is dedicated

    Yes DSL is slower than cable, but DSL connections have dedicated bandwidth (or should), whereas the DOCIS (Data Over Cable Service Interface Specification) based connections provided by cable services use shared bandwidth. IE, with cable, the more active connections to a single DOCIS hub, the slower each is.

    So, in the final analysis, DSL is slower, but more reliable regarding bandwidth and cable is faster, but bandwidth may vary significantly, especially during peak hours.

    1. Anonymous Coward
      Anonymous Coward

      Re: docis (cable) == shared bandwidth - dsl is dedicated

      "DSL connections have dedicated bandwidth (or should), "


      Assuming your DSL provider has abandoned ATM-based backhaul between DSLAM and core ISP network in favour of the cheaper (because shareable) IP-based backhaul, the only dedicated bandwidth in a typical DSL-based broadband network is the twisted pair "last mile" between the DSLAM and the end user.

      The chances of "the last mile" being a limiting factor for most DSL subscribers are quite small. That being said, inevitably some small proportion of unlucky low sync speed "last miles" will be limited by the sync speed rather than by some other bottleneck somewhere else.

  13. AustinTX

    Only the highest speed should ever be seen to come up short sometimes

    Folks DO realize that modems delivering "lower" cable speed tiers are running at the same maximum speed as all the others, but merely rate-limited by software? All an ISP has to do to ensure everyone gets the speed they've contracted for, is to raise the bandwidth cap high enough so that the observed throughput matches the one they're entitled to.

    The same is true of DSL, though they do connect the last mile at varying speeds. However, as I observed with our old DSL service, the configuration is set to a hair's width higher than the contracted speed, resulting in throughput far short of the one we were entitled to. There was no technical reason to do this, as tiers 4x higher were available to purchase.

    In other words, only the very highest speed tier available in the area should ever occasionally fail to deliver the promised "up to" speed, as all lower ones would therefore generally always be achievable. Those data caps are employed inefficiently, resulting in no-one ever seeing the speed they're entitled to.

    Except for cable modem customers. And this is why I seriously question the study. Unlike DSL providers, cable providers like to enable "speed boosts" or periodic "free speed upgrades" due to having so much spare capacity. I've had cable modem many times and often my speed was faster than that which I'd originally contracted for. This cost the cable company nothing, but resulted in my constant satisfaction. But I suppose the rest of the cable providers could be really crappy or stingy....

This topic is closed for new posts.

Biting the hand that feeds IT © 1998–2021