Since its inception, SSIS was meant to execute on a single machine running Windows. The service by itself could not scale on multiple machines. Although it would have been possible to call package execution with custom orchestration mechanism, it didn't have anything built in. You needed to manually develop an orchestration service and that was tedious to do and maintain. See this article for a custom scale-out pattern with SSIS: https://msdn.microsoft.com/en-us/dn887191.aspx.
What lots of developers wanted was a way to use SSIS a bit like the way Hadoop works: call a package execution from a master server and scale it on multiple workers (servers). The SSIS team is delivering a similar functionality in 2017, enabling us to enhance scalability and performance in our package executions.
As mentioned before, the scale out functionality is like Hadoop. The difference...