I'm learning the architecture and design of datawarehouses and have a question on OLAP systems. So as I understand, historically (when computing power and memory weren't a commodity), data would be sourced and stored in a datawarehouse. And for each analytical/BI task, an OLAP system would need to be configured and the query results from the warehouse would be stored in individual cubes which were complex and high-maintenance. Fast-forward to modern age of cloud computing, data is stored either in optimized OLAP cubes, data marts or MPP solutions (columnar stores).
Couple of questions...
Are OLAP cubes virtual/temporary or do they take up physical space on the server? Does data stored in data marts or columnar stores take up the same space as in the datawarehouse? If yes, I assume this isn't an issue now because memory space is readily available.
Is any part of how I understand this wrong? Feel free to correct me, like I said, I'm just learning.
Are OLAP cubes virtual/temporary or do they take up physical space on the server?
->OLAP cubes do take physical space on server. Read about processing cubes. This pulls the source data and stores it in proprietary format.
Does data stored in data marts or columnar stores take up the same space as in the datawarehouse?
-> I think, columnar stores do reduce the space taken compared to row oriented database