what is the most efficient way to get a distinct key list from a big table (aprox 20 mil inserts per minute) ?

equivalent to select distinct key from my_table for this table

CREATE TABLE my_table (

    key text,

    timestamp bigint,

    value double,

    PRIMARY KEY (key, timestamp) )

I need to execute this query quite often ( every couple of minutes )

I can of course maintain a table to hold only unique set of keys but this is of course error prone so I rather avoid it. but it's an option.