Whilst re-reading an article by Bob Frankson on QoS, I was thinking about the difference between what users value from the telephony application (speak to someone not physically present) and what the telcos value (connect devices through a network with 99.999% availability).
The uptime of the application (as opposed to the network) is the chance you can communicate with your intended target given that the target is physically willing and able to receive a phone call. But if they have the TV turned up loud and don’t hear the phone ringing then the network might as well not be there. If they’re out of the house and you phone their landline instead of their cellphone, you get 0% availability. In many ways, a small amount of presence or location information can make a big difference to the “true” application availability. The cost-effectiveness of the last few 9’s of network uptime is questionable. The system is only as good as its weakest link.
The end-to-end perspective is that you have to view the system from the perspective of the user, and not any intermediary. The value of the system is defined by the users at the end points.
Posted by Martin Geddes at 05:49 PMTrackBack URL for this entry:
http://www.telepocalypse.net/cgi-sys/cgiwrap/mgeddes/MT/mt-tb.cgi/13.
Good point. The five-nines criteria for availablity came about as a means for the LECs to measure performance during a time when low-speed data services (dumb pipes) suffered from poorly maintained distribution loops. Nothing wrong with this parameter if you're simply selling dumb pipes.
In an end-to-end system, where the ends are people rather than network element terminations, availability takes on a whole new meaning. What's important to me, on a personal note, is to have the ability to control the 3 Ws - who, when, and where. Without some level of control, I ponder whether my availabilty to others becomes a nuissance.
Posted by: at October 10, 2003 11:13 PMQOS should be in terms of how well you meet customer requirements. Five 9's seems to be a marketing effort by telecom suppliers to encourage the purchase of redundant test equipment. Six Sigma actually beats Five 9's (99.9997% success) and has been incorporated into the manufacturing processes of these same vendors.
The benefit is that it is defined the those factors that are critical to product success. A futher benefit is that a six sigma organization needs less test equipment because the errors occur at a rate that makes reimbursement of lost service cheaper than testing.
This begs the question of why would we (the carriers) go to the expense of devolopling Five Nine's QOS when most commercial service contracts (for customers) state that the maximum exposure of the carrier in the event of a service interuption is the prorata cost of the service? Perhaps the problem lies not in the measure of quality but its cost-effective application.
Network up-time is not a proxy for quality.
Posted by: at October 15, 2003 05:02 PM