We study the problem of process allocation in the course of executing parallel logic programs. Very little work has been done in addressing the process allocation problem in the logic programming paradigm. We observe that the processes that arise in the course of executing AND/OR model programs have very regular and repetitive communication patterns, which justifies the design of tailored dynamic allocation algorithms that utilize such characteristics to achieve better performance. The contributions of this research are: first, addressing the dynamic process allocation problem as a graph embedding problem, where two recently developed graph embedding algorithms are adapted to be used in OM (a virtual machine based on the AND/OR execution model). The performance of the algorithms is evaluated using simulations. Second, developing distributed heuristics for process allocation in OM, one of which is the first algorithm to appear in the literature that combines both sender initiated and receiver initiated strategies. Third, describing a queuing model for OM and analyzing the key parameters of the model and their effect on performance.