MaxCompute - Mars Put Into Commercial Use
Target customers: Python data analysts, data engineers, and developers. Feature released: Mars uses parallel and distributed computing technologies to accelerate Python data scientific computing. It is compatible with NumPy, pandas, and scikit-learn APIs. The Remote API of Mars allows Python functions to be easily and concurrently executed. Mars can be easily integrated with TensorFlow, PyTorch, and XGBoost. If you use Mars in MaxCompute, you can start Mars clusters by using simple APIs. Mars is out-of-the-box, which frees you from creating and maintaining clusters. For example, Mars is automatically installed on DataWorks PyODPS 3 nodes. Mars tasks in a cluster can directly read and write MaxCompute tables. After a Mars cluster is created, tasks can be directly submitted to the Mars cluster for execution. Mars ensures high performance of tasks whose data volume is less than 1 TB.