Same reason we have OLAP and OLTP systems. Same reason it makes sense for some use cases to set up a Hadoop cluster, and for others to have a single but fast CP

Same reason we have OLAP and OLTP systems. Same reason it makes sense for some use cases to set up a Hadoop cluster, and for others to have a single but fast CPU: some tasks can be run in parallel, others can't. See Amdahl's law : https://en.wikipedia.org/wiki/Amdahl%27s_law https://en.wikipedia.org/wiki/Amdahl%27s_law