Dynamics of a Time-outs Management System

A useful tool for distributed systems development is a time-outs manager. A time-outs manager is a software system that manages lists of “timeouts,” that is, objects that postpone a certain function call by a given amount of time. This feature may prove to be useful, as it allows transforming time-based events into nontime-based ones. A typical example is a time-out that on expiring sends an alarm message to the client process. A multiple selection statement such as the C language’s switch can then be used to deal with both classes of events. The focus of this paper is on a time-outs manager that had been designed to facilitate the development of some fault-tolerant applications for European projects TIRAN [1] and DepAuDE [2]. Our timeouts manager maintains a list of time-out objects ordered by ticks-toexpiration as described in [3]. When the specified amount of time elapses for the top of the list, its function is executed and the object is either thrown out of the list or renewed (in this second case, the time-out is said to be cyclic). A custom process checks periodically for the expiration of entries and executes the corresponding functions. This period constitutes a trade-off between performance overhead and maximum function call delay. A key prerequisite for a correct behavior of our system is that the functions associated with the time-outs do not interfere “too much”