At present, there are three modes of Hadoop installation and deployment, namely the following three:
-
Local mode
-
Pseudo-distribution pattern
-
Fully distributed mode
1. Standalone mode (local mode) standalone
In the default mode, there is no need to run any daemon, and all programs are executed on a single JVM. Because it is more convenient to test and debug MapReduce programs in native mode, this mode is suitable for use in the development phase. Use a local file system instead of a distributed file system.
2. Pseudodistributed
Simulate multiple hosts on one host. That is, the Hadoop daemon runs on the local computer to simulate the cluster environment and are independent Java processes.
In this mode, Hadoop uses a distributed file system, and each job is also an independent process managed by the JobTraker service.
The code debugging function is added to the stand-alone mode, allowing to check memory usage, HDFS input and output, and other daemon interactions. Similar to the fully distributed mode, this mode is often used to develop and test whether the execution of the Hadoop program is correct.
3 Fully distributed mode fulldistributed, Hadoop
The fully distributed daemon runs on a cluster built by multiple hosts, which is a real production environment.
Install JDK and Hadoop on all hosts to form a connected network.