Watch this video on YouTube
How does reducing the number of nodes or VMs while keeping the total storage as it is in a cluster affect data shuffling?