The income of telecommunication companies depends on the accurate analysis of customer behavior, on subsequent adaptation of the company to the technological advances in the competitive GSM market, and on marketing new products and services. However, the analysis of sensitive customer data outside these companies is problematic, as it is not usually possible to get informed consent of the customers. In this work, we describe a simulator to generate realistic GSM customer data and open several large datasets generated with this simulator, to the use of researchers. On these datasets, it is possible to test algorithms of customer behavior analysis, including churn analysis, customer segmentation, and anomaly detection.