Spark is 100 times faster than the Hadoop MapReduce in terms of memory and RAM can be utilized to get faster results.
The MapReduce mechanism is very time taking process because the user can write many tasks and tie these tasks using shell/Oozie script.
Always, it is a problem in MapReduce in translating MP output into the input of another MP may require different code. It is because the Oozie script didn’t support.
The user can do everything in Spark by using a single console and can get the output immediately. Switching between ‘Doing something locally’ and ‘Running something on cluster’ is very easy. It leads to more productivity.