Abstract: With the giant data samples authorized by CAIDA, the structural property of Internet topology was analyzed. Sampling biases caused by single point were first presented with the structural characteristics of network topology measured by multiple points analyzed empirically. The results indicated that the real network topology is characterized by power-law distribution and the power exponents decrease gradually with the weakened impact of sampling biases. Then, an analysis is made for the spectral density distribution of network topology to compare the spectral density in the scale-free graph based on BA growth model. It was found that the result is distinctly different from that of a real Internet topology. A conclusion is therefore drawn that the Internet is inconformable to the scale-free graph based on BA growth model.