The 'SORT BY' clause in Apache Pig is used for sorting data in a dataset based on one or more fields. It arranges the data in ascending or descending order, providing flexibility in handling sorted data for further processing.

The default block size used by Hadoop Distributed File System (HDFS) for storing data is 128 MB. This block size is configurable but is set to 128 MB in many Hadoop distributions as it provides a balance between storage efficiency and parallel processing.

Spark is a popular Scala-based tool for interactive data analytics with Hadoop. It provides a fast and general-purpose cluster computing framework for big data processing, making it suitable for various data processing tasks.

The Failover Controller in a Hadoop cluster is responsible for detecting and handling the failure of DataNode machines. It ensures that data availability is maintained by redirecting tasks to healthy DataNodes when a failure occurs.

Apache Sqoop achieves efficient data transfer through parallel processing. It divides the data into smaller chunks and transfers them in parallel, utilizing multiple connections to improve performance and speed up the data transfer process between Hadoop and relational databases.

Replication plays a significant role in ensuring data integrity and availability in a distributed Hadoop environment. By creating multiple copies of data across different nodes, Hadoop can tolerate node failures and maintain data availability.

The Write-Ahead Log (WAL) in HBase is significant for ensuring data durability. It records changes to the data store before they are applied, acting as a safeguard in case of system failures. This mechanism enhances the reliability of data and helps in recovering from unexpected incidents.

The configuration of Hadoop's I/O settings, including I/O buffering, plays a crucial role in cluster performance optimization. Proper tuning can enhance data transfer efficiency, reduce latency, and improve overall I/O performance, especially in scenarios involving large-scale data processing.

The primary role of Apache Flume in the Hadoop ecosystem is data ingestion. It is designed for efficiently collecting, aggregating, and moving large amounts of log data or events from various sources to centralized storage, such as HDFS, for further processing and analysis.

MRUnit's ability to simulate the Hadoop environment is critical for unit testing Hadoop MapReduce applications. It enables developers to test their MapReduce logic in isolation, without the need for a full Hadoop cluster, making the development and debugging process more efficient.