The columnar table is ready, with rows on AMP vertically stored. Partitioning a table helps to reduce I/O by accessing only subsets of columns from a table. It is like creating mini tables from one big table.
To understand this, check the following figure:
The Employee table is stored in a row, so if we need to query a request, all the rows will be read. But if you select particular columns with table partitioned as columnar only, a subset of the table will be read and hence less I/O.
This means that if you select all the columns or more columns from the column partitioned table, it will have lower performance than a normal ...