DIALING BEHAVIOR OF SUBSCRIBERS 441 



accounted for and it is believed that the P.B.X. and coin lines have the same 

 basic characteristics regarding diahng behavior upon encountering busies 

 as have the subscribers who were observed. No significant differences 

 between the results for residential and business offices were noted. From 

 these indirect facts, it is concluded that no significant differences exist be- 

 tween classes of subscribers. 



Effect on the Trunk Plant 



As explained earlier, neither the Poisson nor the Erlang B formula gives 

 an accurate picture of the facts when trunk shortages occur on trunk groups 

 handling subscriber-dialed calls. In both formulae it is assumed that only 

 one attempt is made per call. In the case of the Poisson formula, the call is 

 assumed to be held by the dial equipment until a trunk becomes available or 

 until the subscriber hangs up, and in the case of the Erlang B formula, the 

 call is assumed to clear out. The data developed from the service observa- 

 tions, concerning the dialing behavior of subscribers when encountering 

 busies, indicate that subscribers usually make many subsequent attempts 

 when a busy is encountered. Also the dial equipment with which we are 

 famiUar clears out the calls by giving an all-trunks-busy signal. In order to 

 determine what a trunk capacity table might be like that takes into account 

 the habits of subscribers and the limitations of the dial equipment a study 

 based on simulated traffic was made. This study consisted of 150 CCS 

 (hundred call seconds per hour) of traffic offered to a trunk group varying 

 from 5 to 12 trunks. This study utilized the data developed from the 

 service observations. 



A study based on simulated traffic is a method used to study the capacities 

 of trunking arrangements where a formula is not available. This type of 

 study is based on the idea that calls are placed at random, that holding times 

 of the calls follow an exponential law, and that these characteristics can be 

 simulated by random numbers drawn from an appropriate source. 



The study of 150 CCS of simulated traffic was based on 1,000 calls offered 

 to a trunk group during a ten-hour period. The average holding time per 

 call was 150 seconds, with the total holding time being 150,000 seconds or 

 41.66667 hours. Sub-divisions of an hour were expressed in decimal terms, 

 the smallest division being a hundred-thousandth part. Three sets of 

 random numbers were used for the following purposes: 



1. To determine at what time in the ten-hour period a particular call is 



offered to the trunk group. 



2. To furnish the holding time of a particular call. 



3. To define for each call the pattern of resubmission of the call to the 



trunk group should an all-trmiks-busy be encountered by the call. 

 In each instance the numbers were taken from the tail-end portions of 



