Baidu MapReduce cluster provides four types of nodes:

• Master node: Cluster management node, which is used to deploy NameNode, ResourceManager, HMaster, and other processes, enables high availability mode as needed during the creation of a cluster, and automatically migrates the fault of core component service. The node with more than 4 cores is recommended for configuration.
• Core node: Computing and data storage node, which is used to deploy DataNode, NodeManager, RegionServer, and other processes. All data in HDFS is stored in a Core node. For data security, the core node does not support capacity reduction at present.
• Task node: Computing node, which is used to supplement core the node’s hashrate and deploy NodeManger, PrestoWork, and other processes. No data is saved in node, and computed data comes from the Core node and BOS. Hence, the Task node can reduce and expand capacity at any time.
• Client node: Independent cluster management node and step submission node. It is a group of Baidu Cloud Compute BBC in the same VPC as the cluster, and it provides the user with independent computing resources to realize load balance and security isolation of clusters, improving the cluster’s stability and security.