THEORIES FOR TOLL TRAFFIC ENGINEERING IN THE U. S. A. 505 



In the second solution, 12 high usage and 12.1 common final, or a total 

 of 24.1, trunks were again required, to give 0.03 service to the poorest 

 service parcels of offered load. The overall service here, however, was 

 0.99/91.21 .= 0.011. Thus, with the same number of paths provided, 

 in the second solution (high usage arrangement) the overall call loss was 

 40 pes cent larger than in the first solution,* However, it may well be 

 desirable to accept such an average service penalty since by providing 

 high usage trunks for the first routed traffic, the latter's service cannot 

 be degraded nearly so readily should heavy overloads occur momentarily 

 in the other parcels of traffic. 



9. CONCLUSION 



As direct distance dialing increases, it will be necessary to provide 

 intertoll paths so that substantially no-delay service is given at all times. 

 To do this economically, automatic multi-alternate routing will replace 

 the present single route operation. Traffic engineering of these compli- 

 cated trunking arrangements will be more difficult than with simple 

 intertoll groups. 



One of the new problems is to describe adequately the non-random 

 character of overflow traffic. In the present paper this is proposed to be 

 done by employing both mean and variance values to describe each par- 

 cel of traffic, instead of only the mean as used heretofore. Numerous 

 comparisons are made with simulation results which indicate that ade- 

 quate predictive reliability is obtained by this method for most traffic 

 engineering and administrative purposes. Working curves are provided 

 by which trunking arrangements of considerable complexity can readily 

 j be solved. 



A second problem requiring further review is the day-to-day variation 

 i among the primary loads and their effect on the alternate route system's 

 I grade of service. A thorough study of these variations will permit a re- 



I evaluation of the service criteria which have tentatively been adopted. 

 j A closely allied problem is that of providing the necessary kind and 

 [ amounts of traffic measuring devices at suitable points in the toll alter- 

 ! nate route systems. Requisite to the solution of both of these problems 

 ! is an understanding of traffic flow character in a complex overflow-type 



I * The actual loss difference may be slightly greater than estimated here since 

 i in the first solution (complete access final trunks), an allowance was included for 



i j return attempts to the final route by first routed calls meeting an 0.03 loss, while 

 1 in the second solution (high usage group for first routed traffic) no return at- 



i| tempts to the final route were considered. These would presumably be small since 



I I only 1 per cent of all calls would overflow and most of these upon retrial would be 

 ij handled on their respective high usage groups. 



