In order to improve the real-time performance of Brillouin optical-time domain analyzer (BOTDA), a method based on compressed sensing is proposed in this paper. It contains sparse representation, random sampling and signal reconstruction. Firstly, the method obtained the sparse representation of Brillouin gain spectrum (BGS) by k-means singular value decomposition algorithm, and then BGS can be reconstructed successfully with Gaussian random matrix and orthogonal matching pursuit algorithms. To verify the performance of the proposed method, BGS at different SNR were generated and a 45 km BOTDA was built for temperature experiments. Simulation and experiments show that the reconstructed SNR increase 6.37dB when the average times is 100, which is better than 10.13dB at 3000 average times. We can conclude that the measurement time was reduced 1/30. Besides, the correlation coefficient between reconstructed BGS with scanning step of 8MHz and experimental BGS with 4MHz is 0.9992, which makes the sweeping time decrease 1/2. The compressed sensing method not only ensures the measurement accuracy but also improves the real-time performance of BOTDA.