I have been reading through the Teradata 14 documentation on hybrid columnar structures and its uses in query patterns. I was trying to understand it from the system internal storage perspective and tried to get the understanding of teradata query optimizer in different cases with filters.
I have shared one of the file with my assumptions of representation of data in Columnar Partitions (CP1 to CP5) and Row partitions (RP1 to RP2) in the below link.
Can any one see if my assumptions are right and are representing the data correctly for different query patterns. I understand internal storage is not exactly in the format i showed but will be logically grouped together in the database level and stored in different containers etc.
I am looking for understanding the below special case of table creation on the TD 14 documentation.
CREATE TABLE sales_2 ( store_id INTEGER NOT NULL, product_id INTEGER NOT NULL, sales_date DATE FORMAT 'yyyy-mm-dd' NOT NULL, total_revenue DECIMAL(13,2), ( total_sold INTEGER, top_salesperson INTEGER, note VARCHAR(256) ) ) NO PRIMARY INDEX, -- This clause is optional. PARTITION BY (COLUMN, -- Defines 4 single-column partitions and -- 1 multicolumn partition plus 2 -- additional column partitions for internal -- use. RANGE_N(salesdate BETWEEN DATE '2007-01-01' AND DATE '2013-12-31' EACH INTERVAL '1' MONTH) ADD 36);
I am also looking if Teradata has increased their preferrable limit on maximum number of paraller load utilities (fastload, multiload, fast export) as the new hardware seems to be more faster. We are currently having it set to maximum of 15 paraller load utilities in any time.