THEORIES FOR TOLL TRAFFIC ENGINEERING IN THE U. S. A. 471 



several of its OST alternate rovite groups (similar to those shown in 

 Table V for the Murray Hill-6 office, but not identical) by means of an 

 electromechanical switch-counter having a six-second cycle. During 

 each hour's observation, numbers of calls offered and overflowing were 

 also recorded. 



Although the loads offered to the corresponding direct trunks which 

 ()^'erflowed to the OST group under observation were not simultaneously 

 measured, such measiu'ements had been made previously for several 

 hours so that the relative contribution from each direct group was 

 closely known. In this way the loads offered to each direct group which 

 produced the total arriving before each OST group could be estimated 

 with considerable assurance. From these direct group loads the character 

 (mean and variance) of the traffic offered to and overflowing the OST's 

 was predicted. The observed proportion of offered traffic which over- 

 flowed is shown on Fig. 30 along with the Equivalent Random theory 

 prediction. The general agreement is again seen to be fairly good al- 

 though with some tendency for the ER theory to predict higher than 

 observed losses in the lower loss ranges; perhaps the disparity on in- 



(n) 



0.5 

 0.4 



0.3 



0.2 



0.1 

 



OST N0.1 



THEORY OBSD 



AVG 

 VAR 



2.00 2.36 



5.50 6.52 



RANDOM TRAFFIC 



^--NEGATIVE BINOMIAL 

 -THROWDOWN 



OST NO. 2 



THEORY OBSD 



AVG 

 VAR 



2.10 

 5.60 



2.05 

 6.36 



>RAND0M TRAFFIC 



THROWDOWN 



-NEGATIVE BINOMIAL 



10 15 5 



n = NUMBER OF SIMULTANEOUS CALLS 



15 



p^n 



-NEGATIVE BINOMIAL 



-THROWDOWN 



-NEGATIVE BINOMIAL 



THROWDOWN 



10 15 5 



n = NUMBER OF SIMULTANEOUS CALLS 



15 



Fig. 29 — Distributions of loads overflowing from first alternate (OST) groups; 

 negative binomial theory versus throwdown observations. 



