A preliminary understanding of Hadoop

What the hell is 1:hadoop?

He is a solution, a distributed processing solution that can handle large amounts of data, and is a copycat derivative of Google.

It is the use of Google published MapReduce paper writing into models and frameworks. He mainly divides large tasks into smaller tasks, and gives these small tasks to a single point of execution on the cluster.

What is job, in MapReduce, an application that prepares to commit execution is called a job (job, like a project), and the job, too large, is split into N, executed on each node of the computer, which is called a task.

The Distributed File System (HDFS) provided by Hadoop is mainly to handle storage on each node and achieve high throughput data compilation.

Simply put, it's a resource's storage, and a resource's lookup.

Hadoop uses a master/Master/slave architecture for distributed storage and distributed computing. There is a series of backstage (Deamon) programs. Different background programs play different roles, these roles: Namenode Secondarynamenode,jobtracker,tasktracker,datanode, these names, as long as you touch Hadoop will see, On the master node, the main ones are Namenode,secondarynamenode,jobtracker, and the slave node is mainly made up of datanode,tasktracker.

The

         master node depends on the size of the system for different deployments. When Master is large, the Namenode and Secondarynamenode nodes in master, and Jobtracker allocations are deployed on two servers.