Streaming dead again.


  > Now, back to the NANOG-ish content: I know a fundamental change in
  > technology when I see it, and VOIP is an obvious winner. VOIP has
  > been smoldering for a few years, and the sudden growth of various
  > easy-to-implement SIP proxies and service platforms, plus the sudden
  > drop in price of SIP hard-phones, is going to push growth
  > tremendously. Currently, the underlying technology is UDP that moves
  > calls around. This is all well and good until you get thousands,
  > tens of thousands, hundreds of thousands of calls going at once. QoS
  > is, as Bill says, not a problem right now on public networks; I've
  > used VOIP across at least three exchange or peering sessions (in each
  > direction, no less!) and suffered no quality loss, even at 80kbps
  > rates. However, when a significant percentage of cable and DSL
  > customers across the country figure this technology out, does this
  > cause problems for those providers? Is it worthwhile for large
  > end-user aggregators to start figuring out how they are going to
  > offer this service locally on their own networks in order to save on
  > transit traffic to other peers/providers? Or is this merely a tiny
  > bump in traffic, not worth worrying about?

  > More interestingly: what happens to the network when the first
  > "shared" LD software comes into creation? Imagine 1/3 (to pick a
  > worst-case percentage) of your customers producing and consuming
  > (possibly) 80kbps of traffic for 5 hours a day as they offer their
  > local analog lines to anyone who wants to make local calls to that
  > calling area.

  > Overseas calling I expect will show similar growth. Nobody wants to
  > pay $.20 or even $.10 per minute to Asian nations, so as soon as Joe
  > User figures out how this VOIP stuff works, there will be (is?) a
  > tendency for UDP increases on inter-continental spans. Nothing new
  > here; we've all said this was coming for years. Now it's finally
  > possible - is everyone ready?

  > JT


VOIP is likely to cause a financial upheaval in the telecom industry,
because the overwhelming fraction of revenues still comes from voice
services. However, VOIP is likely to have only a minor impact on
Internet backbones. The reason is that there simply isn't that much
voice traffic. Various estimates (such as those in my papers at
<>) say that already
there is about twice as much US Internet backbone traffic as US long
distance voice traffic, and that is if you count voice as two 64 Kb/s
streams of data. If you use compression, that goes down even further.

Now introducing flat rate VOIP service will stimulate voice usage
some, but based on various previous experiences, not by enough to
make a quantum difference, especially since (again, based on previous
experiences) it will take a while for VOIP to spread widely.

Andrew Odlyzko

About five years ago, before Southern Cross came live and we were struggling to find trans-Pacific bandwidth to New Zealand, we looked at the idea of running our internal voice and IP traffic between NZ and the US on some kind of converged network, to take advantage of the fact that the IP peak load and the voice peak load were about eight hours out of phase.

There were lots of nice graphs that showed a big trough in voice network utilisation almost exactly corresponding with peak IP demand, and everything looked very promising until you noticed that the Y axis on the voice graph was measured in k, and that of the IP graph was measured in megs. The benefit to be gained by being able to burst into the voice trunks was so marginal that it wasn't worth spending the time thinking about how we would do it.

(Every discussion involving mixing voice and data at that company always wound up involving ATM, too, which was another good reason to back away and quietly kill the idea before any madness ensued. I hear it didn't work, though; the company in question was happily running ATM over trans-pacific STM-1s after I left, with AAL5 frames intermingled with circuit emulation. Presumably the 30% cell tax and frame-padding overhead is some kind of ritual offering to the God of "QoS", that magic deity whose name was always invoked to explain why ATM was being used for anything).

If that experience is representative of today's network as a whole, voice is not going to add much traffic to the Internet, relative to traffic that is already carried.

Of course this has nothing to do with whether the Internet today is suitable as a transport for isochronous voice services. But it's always fun to recount an anecdote in which you laugh at ATM.