Each cell in the column is store one after another in memory, you have continuous memory of one column. Effectively think of having text file for each column, each cell having a line, to reconstruct a row you have to get its line from every column file. To sum a column you only have to scan the one file for that column.
In row oriented, each cell for each row is next to each other in memory. Think traditional CSV files, each row is a line in the file. To sum a column you have scan the whole file and all data for all columns.
In row oriented, each cell for each row is next to each other in memory. Think traditional CSV files, each row is a line in the file. To sum a column you have scan the whole file and all data for all columns.