Analysis of large-scale data is challenging due to data storage and computational complexity. When analyzing large-scale data, subsampling methods and divide-and-conquer procedures are appealing, because they ease the computational burden while preserving the validity of inferences. In this talk, several challenges and issues will be reviewed, and a perturbation subsampling approach will be presented based on independent and identically distributed stochastic weights for analyzing large-scale data. The method can be justified based on optimizing convex objective functions by establishing the asymptotic consistency and normality of the resulting estimators. The method simultaneously provides consistent point and variance estimators. We demonstrate the finite-sample performance of the proposed method using simulation studies and a real-data analysis.