Broadband Bottlenecks Come to a Boil

One of the hottest issues in telecom today – how to cure bottlenecks in broadband networks without breaking the bank – is coming to a boil this week on Light Reading.

For starters, April's Research Poll is beginning to deliver some interesting results:

  • Most respondents think the installed based of DSL lines will double in the next one or two years, exacerbating bottlenecks.
  • 19 percent of respondents identify the ATM infrastructure upstream of DSLAMs as the biggest bottleneck.
  • A large proportion of respondents expect carriers to replace ATM infrastructure with Ethernet in the next five years.
  • 55 percent of respondents also expect IP routing to end up beng collocated with DSLAMs in the long run.

    You can take the poll yourself and get an up-to-date reading of the results by clicking on this link.

    Trends such as these are encouraging equipment vendors to roll out next-generation broadband remote access servers (B-RAS), the subject of a free Light Reading Webinar scheduled for this Thursday. Get more information or register for this live event by clicking on this link.

    This Webinar follows from two other recent events covering broadband bottleneck issues:

  • A Light Reading Webinar titled "Upstream of the DSLAM: Beating Broadband Bottlenecks," given by Geoff Bennett, director of Light Reading University, with speakers from LM Ericsson (Nasdaq: ERICY) and the Metro Ethernet Forum. The archive can be accessed by clicking on this link. The webinar previews a report that's just been published on Light Reading (see Upstream of the DSLAM).

  • A panel discussion on "CeBIT Debate: ATM Versus Ethernet" moderated by yours truly at the CeBIT tradeshow in Hannover, Germany. Speakers were from Alcatel SA (NYSE: ALA; Paris: CGEP:PA), Ericsson, Hatteras Networks, Nortel Networks Corp. (NYSE/Toronto: NT) and Siemens Information and Communications Networks Inc. A video archive can be viewed by clicking on this link.

    — Peter Heywood, Founding Editor, Light Reading

  • Page 1 / 2   >   >>
    steviescmr 12/5/2012 | 12:16:59 AM
    re: Broadband Bottlenecks Come to a Boil say, did BobbyMax take this poll yet? you can call it research, but it isn't valid when you allow every Joe LR reader to take this poll.
    sevenbrooks 12/5/2012 | 12:16:57 AM
    re: Broadband Bottlenecks Come to a Boil
    I am amazed that people think that the biggest cost point of DSL is the transport. Given what a DSL modem and DSLAM costs per port, those costs far outweigh the cost of transport. Given that people put up to 10,000 customers on an OC-3c, transport is not the big issue here.

    Erricson has no DSLAM offering so is talking about the piece of the network that they own. But the transport is less than $10 per port. The CO and CPE modems are $125 - $175 per port (added together) at large carriers. Even if Transport was free, the cost change would be less than 10%.

    rtfm 12/5/2012 | 12:16:46 AM
    re: Broadband Bottlenecks Come to a Boil I don't know. Opex costs are higher than capex, in most models I've seen.

    Let's try out some numbers:

    Assume the companies are able to stat mux at a ratio of, say, 10:1. Given uplinking the meg will cost at least $100 per month (normalizing per meg), then we're talking significant uplinking costs. If they stat mux to higher, it helps, but I wager the uplinking bandwidth is at least $80/month per megabit, possibly much more. Regardless, a ball part of $10/month is reasonable, if only mildly high.

    Capex - CPE is a major cost today, esp. until cheaper modems come out. DSLAM/CO costs are modest, when amortized with a 10% discount rate. Of course, one catch remains churn, how long they keep the customer. CO costs can be reused. But marketing, customer acquisition costs, etc. are very high. Reports indicate that with greater competition, the average user contract is falling to perhaps 3.5 years.

    My own $.50


    (total opex > capex on a monthly basis, but yes, transport is not alone the major cost).
    gbennett 12/5/2012 | 12:16:44 AM
    re: Broadband Bottlenecks Come to a Boil I was interested in the strong response that routing would move into the DSLAM. I wondered about the thinking behind this opinion. As many regular posters will know, I come from a strongly pro-ATM background, but I'm trying hard to retain a balance now I'm independent of manufacturer bias ;-) So...

    - Let's consider traffic patterns.
    - Let's assume that P2P traffic makes up between 60% and 80% of residential volume (after all, there's no other reason to have broadband at home right now).
    - Most P2P studies seem to show that you are much more likely to be downloading that MP3 or AVI file from somebody in a different LATA or even a different country than somebody on the same DSLAM.

    To me, these assumptions lead me to think that backhauling traffic beyond the DSLAM before "sorting it out" in the router doesn't impose much of a penalty.

    However, putting a router into the DSLAM does impose a few penalties:

    - It adds cost to a device that's already much too expensive for the current business case.

    - It adds complexity to a device that really needs to be as simple as possible, if not completely passive from a connection management point of view.

    - The current DSLAM manufacturers are not known for routing expertise, so how good will their implementations be?

    - Let's face it, IP already has a lot to prove in terms of "carrier grade" reliability. Just how good are the remote management tools for IP that would allow a complex device like a router to be installed so far out in the carrier network without resulting in an OAM nightmare?

    - Finally, and this is where my conspiracy paranoia starts to show, routing in the DSLAM may open up regulatory issues. While the latest FCC ruling frees up packet-based broadband from unbundling, there's no guarantee that this will be completely replicated in the rest of the world, especially when business voice traffic moves to packets. There may be a move to "unbundle" packet voice service access at the exchange. If DSLAMs continue to be dumb this threat to incumbent domination of broadband services could never happen. For this reason I suspect there'll be pressure by incumbent carriers against DSLAM manufacturers not to introduce routing into the DSLAM. These days the threat doesn't have to be all that strong - just "we won't buy these from you if you put routing capability in there"

    Peter Heywood 12/5/2012 | 12:16:42 AM
    re: Broadband Bottlenecks Come to a Boil Yesterday, with 24 responses, the biggest bottlenecks identifiedl were:

    1. Server (33%)
    2. Internet peering points (24%)
    3. ATM infrastructure (19%)

    Now, with 77 responses, we have:

    1. ATM infrastructure (37%)
    2. Server (19%)
    3. Internet peering points (18%)

    I guess this proves how influential a little article in Light Reading can be!
    Peter Heywood 12/5/2012 | 12:16:42 AM
    re: Broadband Bottlenecks Come to a Boil Just wanted to point out that Geoff Bennett's report on the design issues influencing technology choices upstream of the DSLAM was posted yesterday:

    Fiber_me_up 12/5/2012 | 12:16:42 AM
    re: Broadband Bottlenecks Come to a Boil Sorry to ask a basic question on a fairly complex issue....but if the peaks in consumer traffic demand are smaller and occur at different times than for enterprise/business traffic demand, isn't the consumer traffic getting a free ride? Therefore, why all the concern about congestion in the backhaul?


    gbennett 12/5/2012 | 12:16:41 AM
    re: Broadband Bottlenecks Come to a Boil In the past, residential traffic happened in the late afternoon and early evening as kids got home from school, and parents from work. This happened just as business traffic throttled back because everyone goes home.

    I haven't seen data to back this up, but if we're moving to a P2P model, then I assume there'll be a constant background level of traffic for residential broadband 24 hours a day because the transfers can take days (for a movie), and will continue when the PC is unattended.

    In contrast, database updates and off-site backups kick off at night when the network is "quiet".

    In fact I suspect there's evidence of both these effects, but they show up as a raised average utilisation level, rather than a disappearance of the time-dependent peaks.

    In terms of backhaul congestion, remember that business and residential traffic may not share the same link until deeper into the network (where bandwidth tends to be more freely available). This is partly because we often put houses and businesses in different places (they are mixed sometimes too), but it's also because carriers will deliberately put business traffic onto a separate physical or virtual network so they can control the level of over-subscription on that network, and so ensure a lower level of congestion to the business subscriber.

    IPee 12/5/2012 | 12:16:41 AM
    re: Broadband Bottlenecks Come to a Boil The DSLAM can give the service provider more control. They can for example limit a single person P2P traffic more easily because they know a particular line is to a particular subscriber without having to do any layer 2 or layer 3 checks.

    Although you can use the DSLAM as a control point, I am less certain that the routing will move there. However, I would expect the DSLAM to start to play a more active role.
    sevenbrooks 12/5/2012 | 12:16:40 AM
    re: Broadband Bottlenecks Come to a Boil

    Ummm.....ok, I agree Opex > Capex.

    But my view of Opex for Transport is that is approximates 0. If you are a LEC then the Opex cost is that of having another blade in your local SONET mux. That is why I was discussing Capex.

    I would think the Opex associated with loop management would be much greater than that associated with transport (line issues, churn, etc.)

    Can you describe to me where my thinking is off?

    Page 1 / 2   >   >>
    Sign In