Hadoop gen1 vs Hadoop gen2

In YARN (the new execution framework in Hadoop 2), MapReduce doesn't exist in the way it did before.

YARN is a more general purpose way to allocate resources on the cluster. ResourceManager, ApplicationMaster, and NodeManager now consist of the new YARN execution framework. The NodeManager is the daemon on every node, so I guess you could say that replaced the TaskTracker. But now it just gives processes instead of just map tasks and reduce tasks.

MapReduce is still there, but it is now an "application" of YARN.

Here is an introduction to YARN, which will go into much more depth: http://hortonworks.com/blog/introducing-apache-hadoop-yarn/

Yes Jobtracker was split into resource manager and application master.Application master runs on one or all node managers instances based on the number of jobs submitted.So when job submitted, resource manager talks to one of free node managers to act as application master and that application master will be now job tracker and other node managers will be task trackers which they execute Yarn child.Correct me if I'm wrong.

What I get after reading above link is

YARN handle the shortcomes of classic MR by splitting the functionality of Job tracker

functionality of JobTracker in 1.x i.e resource management and job scheduling/monitoring are divided into separate daemons. - global ResourceManager (RM) and per-application ApplicationMaster (AM)

ResourceManager - run at NameNode i.e master side

• it DISTRIBUTE RESOURCES among all appl

  it has 2 main components: Scheduler and ApplicationsManager.

• Scheduler is pure scheduler
• ApplicationsManager is responsible for accepting job-submissions

NodeManager - run at DataNode i.e slave side

• is the per-machine framework agent
• it is responsible for containers, monitoring their resource usage (cpu, memory, disk, network) and reporting the same to the ResourceManager/Scheduler.

Central ResourceManager and Node specific Manager together is called YARN