We present analytical and numerical results of modeling of flows represented as correlated non-Poissonian point process and as Poissonian sequence of pulses of different size. Both models may generate signals with power-law distributions of the intensity of the flow and power-law spectral density. Furthermore, different distributions of the interevent time of the point process and different statistics of the size of pulses may result in 1/f ^β noise with 0.5 ≲ β ≲ 2. A combination of the models is applied for modeling Internet traffic.