While the c4.8xlarge instance will be preferred for the master machine, the i2.large instance seems fit for the slave machine. Another way is to launch the Amazon EMR instance that automatically configures the servers.
Hence, you need not deal with manually configuring the instance and installing Hadoop cluster while using Amazon EMR instance. Simply dump the data to be processed in S3. EMR picks it up from there, processes the same, and then dumps it back into S3.
