Load allocation is the resource management component in the distributed system. Its aim is to schedule the system resource to optimize the system performance. Generally, load allocation is a NP-complete problem. There are practical values to solve the problem using random method. By spectrumizing the problem to study statistical features in the process of load allocation, the paper focuses on the changes of standard deviations in all kinds of random graphs and provides theoretical evidences for dependable software reuse. Three research topics are covered: (1) Using random graph with Poisson distribution to study the relations between vertexes in multiple-choice-paradigm. Studying the load standard deviations of multiple-choice-paradigm in directed random graph and utilizing the generation process of random graph to extend the classical multiple paradigm, and the physical meaning of 1 |