Some researchers used the conception of phase transition to explain ubiquitous power laws in the complex Internet. However, this explanation obviously ignores the correlation between nodes' behaviors. In this paper, we compare the phase transition with the collective correlation behaviors of nodes by use of a computer network model. It is shown that when the load of network is increased, the spatial and time correlation between nodes'behaviors is also strengthened and the collective correlation phenomenon emerges gradually. At the critical state, the system load fluctuates with a 1/f2 power spectrum, and the power spectra of buffer contents fluctuations vary as 1/fα noise where α≤2. The existence of spatial-temporal correlation among nodes'behaviors is further demonstrated.