ErlangB relates to a static, wired network where you “block” call originations when your existing trunk group does not have capacity to carry them based on its current calls in progress and their duration.
It is an estimate only, and most useful for trying to determine if you have enought capacity for the largest block of calls you are likely to expect (the BusY Hour Busy Day concept). The allowed blocking rate at that point would be reflected as a P, followed by two digits. The “P” presumably stands for “Point” pr “Period”, and the two digits represent how many calls blocked per hundred attempts. This goes back to a time when you had only teletypes and typewriters to contend with, and the “P” replaced the period that would easily go unnoticed. Thus, .03 became P03, and meant that at most, 3 calls out of a hundred would be blocked at the Busiest Hour of the Busiest Day of the year, which is traditionally 2 PM on Mother’s Day, when all the kids want to call and talk to Mom.
But all that becomes much less relevant when you have your networks capable of carrying tracket on alternate circuits, when you have Call Waiting, when you see more and more computers using dialup services and staying on line for longer periods of time, when you find that Data Communications increasingly rely on continuous connnections (no waiting for dialtones anymore), and when you consider Voice Over IP (VoIP) and wireless services where voice is carried in a string of packets via cell technology rather than with a fixed connection.
The fact is, trying to fit the ErlangB model to those changes without understanding that there are significant distinctions involved, means trying to string oranges on a vine as though they were grapes. You might be able to succeed, but will it be what you think it is when you are done?
To put it another way, a good algorythm becomes a bad one when it is used with the wrong data or applied in the wrong manner.