YARN's Multi-Tenant mode enables the running of multiple workloads simultaneously on a shared cluster. It allows different applications to share cluster resources efficiently, supporting a diverse set of workloads.

Batch processing jobs in Hadoop are typically scheduled using Apache Oozie. Oozie is a workflow scheduler that manages and schedules Hadoop jobs, providing a way to coordinate and automate the execution of complex workflows.

In MapReduce, the process of breaking down large datasets into smaller, manageable chunks for individual mappers is called Input Splitting. These splits are then processed in parallel by the Mapper tasks to achieve distributed computing and efficient data processing.

When dealing with sensitive data, data encryption becomes a crucial aspect of security in Hadoop. Encrypting data at rest and in transit ensures that unauthorized access is prevented, providing an additional layer of protection for sensitive information.

In a scenario of data loss, NameNode is a crucial Hadoop process to examine for any potential recovery. The NameNode maintains metadata about the data blocks and their locations. Understanding the state of the NameNode is essential for data recovery strategies in case of failures or data corruption.

When facing 'Out of Memory' errors in Hadoop, it's crucial to inspect the 'mapreduce.map.java.opts' configuration parameter. This parameter determines the Java options for map tasks and can be adjusted to allocate more memory, helping to address memory-related issues in MapReduce jobs.

Sqoop can streamline the process of migrating legacy data from multiple databases into Hadoop by using the Multi-table Import functionality. It enables the concurrent import of data from multiple tables, simplifying the migration process and improving efficiency.

Impala is known for its ability to perform real-time queries on Hadoop. It is a massively parallel processing (MPP) SQL query engine that delivers high-performance analytics on large datasets stored in Hadoop. Unlike traditional batch processing, Impala allows users to interactively query and analyze data in real-time.

The Hadoop Distributed File System (HDFS) is designed to store and manage large datasets across multiple nodes in a distributed environment. It provides fault tolerance and high throughput for handling big data.

To optimize performance in Hadoop data pipelines, shuffling techniques are employed for effective data partitioning and distribution. Shuffling involves the movement of data between the Map and Reduce tasks, facilitating parallel processing and efficient resource utilization.