3/18/2005
When is good enough not good enough?
Larry Borsato has posted an excellent observation that seems to continually escape the VoIP proponents who are so enthused with the technology that they can’t embrace the reality.
Mark Evans and Ronald Gruia are both wondering just how good Voice over IP (VoIP) phone service has to be to reach mass acceptance. They both believe that customers will be willing to live with less that 99.999% (”five nines”) reliability, if the price is low enough.
To put this into perspective, 99.999% reliability means that phone service may potentially be unavailable for 5.3 minutes over the course of a year. By comparison, 98% reliability sounds reasonable, but actually the service could be unavailable for more than 7 days.
Could you live without phone service for hours or days at a time? Probably not.
That’s a key difference brought about by 100 years of engineering maturity. One of the benefits the PSTN has over the Internet I’ve always been vocal about is traffic engineering. The PSTN was designed to support a single type of traffic (voice) very, very well. The Internet was designed to support all types of traffic (packetized in IP) using best efforts over available resources. This is one difference that is not well understood.
And that still requires further analysis. Do you personally feel you get 98% uptime from your ISP?
As to why telcos would cannibalize their own revenue stream, they’ve always done that. And in many way (see his comments on cellular) we’ve lowered our standards to accept less that 5 nines reliability as the norm. How do we quantify the impact of accepting lesser quality or availability?
Filed by Ken at 11:04 am under General
Add to all that the “fun” that ILECs (and Cablecos) can (and will) have with packet tagging and packet interleaving, and you have a completely impossible situation for any 3rd party VOIP provider wannabees. Vonage may be big now but their days are numbered.
For the uninitiated - packet tagging, a.k.a. multiprotocol label switching (MPLS), places a digital identifier into each data packet originating on that network. The tag identifies the type of packet - such as voice or video for example. MPLS allows the carrier to handle the packets at the switching layer (layer 2 in the OSI) rather than the routing layer (Layer 3). By doing this carriers can assign a higher priority on the network to one set of tags over another. Alert readers can see where that takes you, assign a higher priority on your network to your own tagged packets than those from 3rd party providers thereby making “best effort” on the network for those 3rd parties much worse than before.
Packet interleaving refers to the way packets are sent over a network. To overcome problems of packet loss (packets that get corrupted or do not make it from source to destination) packets are generally sent multiple times. Interleaving is the sequence of those packets. If you were to number the packets 12345 and send each packet 3 times 111222333444555 you stand a greater chance of losing one set of packets than if you interleave them 123542354513421. This opens a possibility for carriers to do some nasties with how they pack the VOIP packets of 3rd parties like Vonage. Bob Cringely talked about this very thing this week (http://www.pbs.org/cringely/pulpit/pulpit20050317.html) - the carriers could bunch all competitors VOIP packets into the first half of every second thereby leaving “dead air” as it were which would be very noticable as a delay in a VOIP call.