I have a table

create table t (
    index_column_a integer not null,
    index_column_b integer not null
    value_column_a integer not null,
    value_column_b character varying(256) not null
    ...
);

Let’s say:

• the table has 100,000,000 entries
• there are 10 distinct possible values of index_column_a
• there are 10,000,000 distinct values of index_column_b
• the combination of index_column_a and index_column_b are unique
• I want to access 100,000 records at a time by specifying:
  • a single value for index_column_a
  • a between range for index_column_b

Which of the following is a more efficient index for select speed?

create unique index t_idx on t(index_column_a, index_column_b);

or

create unique index t_idx on t(index_column_b, index_column_a);