All MCQs

The Scheduler decides which applications receive containers.
It does not directly monitor the internal logic of tasks.

Regular heartbeats carry resource and status information.
The ResourceManager uses it for scheduling decisions.

The ApplicationMaster is the first application-specific process launched.
It then requests additional containers for the workload.

Each YARN application has an ApplicationMaster.
It negotiates resources and monitors the application's tasks.

Containers provide bounded resources in which application processes run.
They are logical allocations rather than virtual machines necessarily.

Each worker runs a NodeManager that launches and monitors containers.
It reports node status and resource availability to the ResourceManager.

The ResourceManager arbitrates resources among applications.
It includes scheduling and application-management responsibilities.

YARN separates cluster resource management from application processing logic.
It became the general resource layer of Hadoop 2.

Fewer, larger splits reduce mapper count and scheduling overhead.
The underlying small files remain separate.

Task startup cost can dominate actual processing for tiny inputs.
Container files or combined input formats can improve efficiency.

Unsplittable compressed files may be assigned to one mapper.
Splittability preserves input parallelism.

Intermediate data can be large and temporary.
Compression often saves more transfer time than it costs in CPU.