Mark Wong <markwkm@xxxxxxxxx> writes: > On Mon, Feb 10, 2014 at 9:20 AM, CS DBA <cs_dba@xxxxxxxxxxxxxxxxxxx> wrote: >> In the case of this being a timestamp I suspect the performance would >> take a hit, depending on the size of your fact table and the >> scope/volume of your DSS queries this could easily be a show stopper >> based on the assumption that the database can do a numeric binary search >> much faster than a timestamp search > I guess I was hoping the extra 4 bytes from a timestamp, compared to a > bigint, wouldn't be too significant yet I didn't consider postgres > might do a binary search faster on an integer type than a timestamp. Actually, Postgres timestamps *are* bigints under the hood, and comparisons will be about the same speed as for bigints. Now, I/O is a lot more expensive for timestamps ... but not index searches. regards, tom lane -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
