≡ Menu

The Application-centric Carrier

For some time now, I’ve been referring to a concept in presentations that I’ve titled “Carrier 2.0”.  An off-shoot of the Voice 2.0 manifesto, the primary concept is that the service provider of the future will be horizontally integrated, not vertically.  It will be a very different beast from the incumbents we know today.

Today’s vertically integrated carriers provide network access, originations, terminations, applications, and identity, all in one neat bundle.   It’s convenient, but it masks the costs of providing those services from the consumer.   However, today, anyone, literally, can be a service provider.  It’s a simple enough formula:

  1. Get yourself an application server.  Asterisk will do nicely, and has the added advantage of being inexpensive, and open, so you can customize it any way you like.  With a little tweaking you can get 300 or more simultaneous callers from a state-of-the-art server.  That’s enough to service 3000 or so customers.
  2. Outsource your terminations and originations.  There are plenty of folks who will supply both to you, and whether your choice is a smaller regional network like Unlimitel, or a large international like XO or Level3, there’s someone who will provide these services to you.
  3. Bill.

Vonage’s Jeff Citron was dealt a nice softball interview last weekend by the NY Times.  A lot of people commented on it at the time, noting that Vonage’s model was in serious question.  Jon Arnold, especially, wrote a lengthy comparison between Vonage and one of his favorite companies, Telio.

What if the world of the Vonage’s and the Telio’s is dying?  What if, like the PC revolution of 30 years ago, the vertically integrated service model gives way to a horizontal model?  Just like the first PCs, hackers using Asterisk are building their own custom phone services.  For example, I myself have a $300 PC in my basement, running Asterisk, and terminating calls on five networks.  In theory, I should be able to handle a few hundred simultaneous callers on that box.  It has got the memory, the CPU power, and the connectivity.  Am I a service provider?  I’ve seriously contemplated setting up accounts for my extended family on this box. 

If a single tweaked Asterisk server can handle three hundred to five hundred calls, then a dual core machine might handle 1000, and the coming generation of quad core machines might handle 2000.  Using ordinary residential 10:1 subscriber to calling ratio models, that 2000 simultaneous callers, should handle 20,000 subscribers.   That, my friends, is an RLEC. 

Detractors point to the many failings of Asterisk — redundancy, call quality (it can suffer if not properly tuned), missing features.  But just as PCs were dismissed thirty years ago by the IT department and ultimately surpassed that centralized command and control infrastructure, so it’s also clear that commodity PC hardware, running Asterisk (or something like it), will also surpass today’s purpose built telecom equipment.

The proof?  Asterisk is showing up in applications everywhere. It’s the dirty little secret of the Fortune 500 — need a voice server?  Asterisk will do the job.  It’s also the foundation of so many of the innovative new startups being funded these days — Jajah, and Radio Handi to name a couple.  It won’t be long before a scalable, carrier class Asterisk solution, built on commodity hardware, is common place.

Mr. Citron’s world is changing.  Tomorrow’s service provider will be a software company.  It won’t be the vertically integrated marketing machine and network operator that he’s built.  It will be a value chain of horizontal relationships. Can Vonage evolve fast enough?  Tellingly, in the entire NY Times article there was a scant single paragraph devoted to future services.  Vonage seems to be stuck in the past.

The service provider of tomorrow probably will target niches –  a particular application, or a particular audience.  Just as today we’ve seen portal players evolve two distinct advertising strategies — MSN and Yahoo focus on selling to large advertisers, while Google’s contextually driven adsense is focused on every niche player — it’s likely we will two categories of service providers emerge: large players offering homogeneous services to a mass audience, and smaller players offering distinct niche offerings to defined smaller markets.  And, over time, we should expect the niche players to grow at the expense of the mass players.

When anyone can be a service provider, is there room for the so called “pure play VoIP” anymore?  Is there room for today’s Vonage?

{ 6 comments… add one }

  • Brad Jones September 6, 2006, 6:37 am

    Great post and great blog! Curious about your statement, "… and terminating calls on five networks." Can you elaborate on which networks and some pricing?

  • Alec September 6, 2006, 6:47 am

    Thanks Brad!

    I terminate on SipPhone (1 cent / minute, North America), NuFone (2 cents / minute, North America), Unlimitel (2 cents / minute, North America), VoIPdiscount.com (free to 49 countries), and my local carrier via a POTS card, Bell Canada.

    I also buy a DID from Unlimitel, in the 613 area code.

    Why so many? Because I can :) Quality varies, naturally, and VoIPdiscount is often flaky. I guess you get what you pay for. Nufone and Unlimitel are both excellent.

  • Eric April 30, 2007, 5:58 pm

    While i do believe that asterisk has the capability to migrate to carrier grade solutions, your numbers are totally off.

    I work at a carrier grade voip switching company. Our proprietary switches can handle 1000's of calls simultaneously. 300 calls will absolutely not cut it. I think you need to scale up vertically a bit, and then rely on horizontal scaling. You must be able to match the call density of a proprietary switch now to entice any use in carrier grade settings.

    Also hardware transcoding, and TDM to VOIP is a big deal. With the newly available DS3 cards, and some hardware codec transcoding cards.. i feel like it is becoming more feasible to use an asterisk box in a carrier setting.

    The bottom line though.. is that if you think running 300 calls (signaling and media) will get you into carrier grade… forget it.

  • Alec May 1, 2007, 2:17 am

    Eric, I know of people doing this, as I said, with quad core gear. Over 2000 simultaneous calls per server, and racks of 2U servers handling the traffic. As for transcoding — agreed. Typically, what they're doing is buying origination and termination from a third party. That way the transcoding is done for them, and the traffic delivered VoIP to their servers.

  • wyane February 5, 2008, 7:01 pm

    Hi Alec,

    Its a good post, as I am a new VoIP engineer and learning VoIP technology.

  • Diriye Barre March 22, 2010, 11:00 pm

    Dear Sirs/Madam,
    March 23/2010.
    This article of Astriks seems, what I was looking for long time and asked many people, but I just read to night March
    23/ 2010. and I very excited.
    Is there anybody who can explain more, and share with me
    building such a network.
    Can we use astriks to connect directly communicating places
    such East Africa, without using carriers?

Leave a Comment