Asynchronous and Archimedean timing

In this timing approach, the steps and the communication delays are randomly chosen from a time distribution. More specificaly, every time a node executes an action of the protocol, selects a new step (and remains idle for this amount of time) depending on the time distribution selected and the time distribution parameters for its step. Also, every time a message is sent, a delay is chosen depending on the time distribution selected and the delay parameters for the communication line used.

A constraint in this concept is that, on purpose to keep the simulation steady, the step of a node can be changed only when no timer is running. This constraint seems hard but in all the simulation experiments with protocols with many timers the steps changed lots of times and no problem arised.

The time distributions available are uniform, 2 types of geometric, normal and deterministic. For every kind of distribution, the parameters required for every node and link are given in the network file.

In the case of asynchronous protocols it is better to select wide limits for the time distributions.

In the case of Archimedean protocols you choose the limits of the distribution of the steps and delays to vary as much as you want. In the simulator the upper and lower step and delay limits for all the distributed system are found and are available for use in the global variables smin, smax and dmin, dmax.

Another point that must be taken care is that the initial times in the network description file must be equal when simultaneous initiation is desired.

SEE ALSO: System structure (Section 5.2), Randomizing routines (Section 7.2.7) and Making new networks (Section 6).