Abstract :
This work attempts to bring queueing, buffer, and network models still closer to their important current applications to information systems and computers by incorporating major features peculiar to these systems. These features are (i) finite capacities, already broached in the theory of dams; (ii) the phenomenon of loading, in which the arrivals of items are not instantaneous but take a time typically proportional to their size; (iii) the need for storage space as well as service time; and most important (iv) the conservation principle that items (such as data blocks, programs, messages, etc.) not change their sizes as they move about in the system. Although recognized as a theoretical requisite for modeling many information systems, the conservation principle has been conspicuously absent from nearly all the major models under recent consideration. In these the successive sizes of an item in its appearances at a sequence of system nodes often form an i.i.d, stream, when they should really be one and the same random variable.
