Wikipedia: Erlang

Difference (from prior major revision) (author diff)

Changed: 1,20c1,4

Erlang is a statistical measure of traffic. It is named after the Danish telephone engineer, A.K. Erlang, the originator of [queueing theory]?.

Commonly used in telephony, it refers to the use of a communication channel, particularly voice and in relation to a PABX, but may also refer to any traffic system.

In the "Erlang-B" calculation, one Erlang implies a single channel in continuous use (or two channels at fifty percent use, and so on, pro rata), usually for one hour. For example, if a bank has two tellers and during the busiest hour of the day they're both busy the whole time, that would represent two erlangs of traffic.

Typically erlangs might be used to determine if a system is over- or under- trunked (has too many or too few phone lines). It might also be used to measure traffic on a T-1, to determine how many voice lines are in use at the busiest hour of the time period being examined; for 24 channels, if only 12 are ever in use, the other 12 might be made available as data channels. The erlang calculation also determines "grade of service" or "blocking factor" - if a user tries to make a call during the busy hour, how likely is it that they will get a busy signal (typically a blocking factor of 5% or 1 in 20 is considered acceptable).

There are a range of different Erlang forumlas, depending on assumptions about caller behaviour when encountering a busy tone.
* Erlang B
* Erlang C
* Engset formula (not sure is this is due to Erlang, but it's related)

It might be nice to put the calculation itself here?

See
* [Erlang formulae] (in Dutch)
* [www.erlang.com] for more information.