Climate dataset is the basis of satellite-based climate research. This paper formalizes some basic concepts of satellite data from the view of climate study first. On that basis, the classification problem of Climate Data Records(CDR) is discussed. The authors proposed that current classification method, which classifies CDR into two categories, FCDR(Fundamental Climate Data Record)and TCDR(Thematic Climate Data Record), is not enough to reflect the characteristics of remote sensing dataset. It's necessary to further classify TCDR into three categories, including dataset generated from single instrument, dataset generated from multiple instruments, and dataset generated from multiple sources. The advantage of this classification scheme is that it's helpful to discriminate climate datasets which have the same name but generated from different data sources or by different algorithms. In this manner, users can understand and utilize the data more easily and correctly. Then, this paper introduces Essential Climate Variables (ECV), essential satellite-based climate variables and the guideline for the generation of satellite-based climate datasets respectively, and reviews the recent progresses made by NOAA, NASA, ESA and other international remote sensing organizations. The author especially emphasized that some guidelines and principles for the generation of satellite-based climate data records have been proposed and made common senses among the majority of international remote sensing organizations, and new climate data records following international standards are being produced. Finally, the paper discusses some issues faced during the construction and application of satellite-based climate datasets, and gives some specific suggestions on the development of China's satellite-based climate datasets.