298 THE BELL SYSTEM TECNHICAL JOURNAL, MARCH, 1954 



No completely satisfactory answers to these questions have yet been 

 found but some knowledge of the answers can be gleaned both from 

 theory and from much experience in the operation of intertoll trunk 

 networks. Let us consider each question in turn. 



Random Versus Non-random Traffic 



Statistical theory states that the portion of random traffic which fails 

 to be carried when offered to a given number of trunks is not itself ran- 

 dom in nature. Such portion is, of course, the overflowed or "lost" calls 

 described in the discussion of the Erlang B formula used in the separa- 

 tion of load between direct and alternate routes. Each such portion 

 should be increased in volume to equal an equivalent random value before 

 being offered to another trunk group since both the Erlang B and Poisson 

 expressions of trunk capacity are predicated on the offering of random 

 traffic. Such adjustment of high usage group overflows was not made for 

 two reasons: one was the absence of any suitable increase factors for 

 the variety of conditions encountered and the other was a definite indi- 

 cation from the New York City trials of alternate routing previously 

 mentioned that such factors were not needed. Analysis of the New York 

 trial indicated that when many parcels of overflow traffic are presented 

 during one hour to the final route the aggregate tends to become random 

 in nature. However, it was deemed advisable to provide some extra 

 trunks in the final route to protect service by absorbing any abnormal 

 peakiness in the offered traffic which might be due to non-random charac- 

 teristics. As a practical matter it was possible to observe and regulate 

 the loading of the final tandem route to the end that a satisfactory over- 

 all grade of service was maintained and from this and other subsequent 

 experience to devise a rule-of-thumb for engineering future requirements 

 in such final routes. 



A similar approach to the problem was adopted in engineering the 

 nationwide intertoll trunk network. An arbitrary increase was applied to 

 the computed load offered to each final group. Whether this correction 

 was too little or too much is still a moot question but it is believed that 

 it was proper in the sense that it was in the right direction. 



Statistical theory, however, does not support the view that a combina- 

 tion of non-random parcels tends to produce a random aggregate. The 

 problem is under further study by the Bell Telephone Laboratories with 

 a view to producing some practical means by which field engineers can 

 readily and with reasonable accuracy adjust non-random trunk overflow 

 traffic to equivalent random values for projection purposes. 



