The task allocation problem in a distributed environment with heterogeneous nodes connected by communication links is treated. In this problem, both the nodes and the communication links are unreliable: each has a certain failure rate. The goal of the task allocation is to maximize the system reliability through an appropriate placement of the tasks. The paper contains a probabilistic reliability model of the system, three heuristic task allocation algorithms, and some analysis of the efficiency of these algorithms in reaching their goal.

The main problem with the model is its validity. One essential assumption in the model states that if a component fails during an idle period, the failure will be detected and repaired immediately (in zero time). This assumption restricts the applicability of the model. The model itself gives some insight into the reliability analysis of a complex system and into the solving of allocation problems. The comparisons between the algorithms, based on “randomly” generated networks, are interesting. The reader should not expect any systematic treatment of the topic, however, only some general characteristics. The presentation is uneven, especially in the choice of material.