The thought of divide-and-conquer of traditional row database
In order to avoid generating a large amount of disk io, the traditional row database divides the data into a datapage, for example, using the maximum length of the row in the 4k datapage datapage (even if only one byte of data is stored), calculates the number of rows that each row can hold, and takes out the index column of each row to generate an index table. When reading the data, you only need to go to the index table to get the datapage where the data is located. In other words, only a 4k io operation was performed.
When the data addition, deletion and modification operation occurs, because the maximum length of the row has been occupied, the change of the size of a row of data is avoided, and the data needs to be migrated at the expense of storage to improve efficiency.
