Several existed models for scalability analysis are introduced. The time-constraint and memory-constraint speedup laws are analyzed from a new viewpoint. Based on these models, we extract the essential of this metric and give a generalized definition for it when the parallel system is composed of a parallel algorithm and an isomorphic architecture. From this definition, present three other models for scalability analysis: equal-average-I/O-requirements model, equal-average-communication-requirements model and equal-utilization model. Finally, We discuss the extension of these models to cluster of workstation systems.