Problem
a) What are the purposes of two master nodes in Hadoop? How does the master achieve high reliability to avoid the single point of failure?
b) What is speculative execution (also called backup tasks) in Hadoop? What problem does it solve?
c) You have a data file (480MB) and want to store it in HDFS which block size is 128 MB. Your Hadoop cluster has 10 data nodes. Explain (draw a figure if needed) how this file is divided and stored in this system in details (e.g., with specific locations of data blocks). And explain the steps how this file can be accessed later.