Hello all,
we hav a large table (CDRs) with PPI defined on timestamp containing date & time when a record was loaded in DW.
CREATE SET TABLE CZ_PRD_TGT.AMAOUT_NA ,NO FALLBACK , NO BEFORE JOURNAL, NO AFTER JOURNAL, CHECKSUM = DEFAULT, DEFAULT MERGEBLOCKRATIO ( AMAOUT_NA_ID VARCHAR(60) CHARACTER SET LATIN NOT CASESPECIFIC TITLE 'Amaout Network Activity Id' NOT NULL, REC_NUM_SEQ INTEGER TITLE 'Record Number Seq' NOT NULL, ...................... /* large number of columns here, not importat right now */ LOAD_DTTM TIMESTAMP(6) TITLE 'Load Dttm' NOT NULL, HIST_TYPE INTEGER TITLE 'Historization Type' NOT NULL COMPRESS 990000005 ) PRIMARY INDEX AMAOUT_NA_NUPI_PPI ( AMAOUT_NA_ID ) PARTITION BY RANGE_N(CAST((ACTV_START_DTTM ) AS DATE AT TIME ZONE 'Europe Central') BETWEEN DATE '2013-02-01' AND DATE '2013-09-30' EACH INTERVAL '1' DAY , NO RANGE, UNKNOWN);
However, because of a clear business requirement, multiple level partitionning would be much better - i.e. first level of partition should be the business validity of a record and the second level the should be the date & time when the records were loaded in DW:
CREATE SET TABLE CZ_PRD_MNC.AMAOUT_NA ,NO FALLBACK , NO BEFORE JOURNAL, NO AFTER JOURNAL, CHECKSUM = DEFAULT, DEFAULT MERGEBLOCKRATIO ( AMAOUT_NA_ID VARCHAR(50) CHARACTER SET LATIN NOT CASESPECIFIC TITLE 'Amaout Network Activity Id' NOT NULL, REC_NUM_SEQ INTEGER TITLE 'Record Number Seq' NOT NULL, ...................... /* large number of columns here, not importat right now */ LOAD_DTTM TIMESTAMP(6) TITLE 'Load Dttm' NOT NULL, HIST_TYPE INTEGER TITLE 'Historization Type' NOT NULL COMPRESS 990000005 ) PRIMARY INDEX AMAOUT_NA_NUPI_MLPPI ( AMAOUT_NA_ID ) PARTITION BY ( RANGE_N(CAST((ACTV_START_DTTM ) AS DATE AT TIME ZONE 'Europe Central') BETWEEN DATE '2013-02-01' AND DATE '2013-09-30' EACH INTERVAL '1' DAY , NO RANGE, UNKNOWN), RANGE_N(CAST((LOAD_DTTM ) AS DATE AT TIME ZONE 'Europe Central') BETWEEN DATE '2013-02-01' AND DATE '2013-09-30' EACH INTERVAL '1' DAY ,NO RANGE, UNKNOWN) );
What we found was that the volume of table with MLPPI was remarkably higher - original table had ~ 1TB, the new table with MLPPI has ~1,3 TB!
Does anybony know what is the reason that MLPPI requires so much higher disk space?
Thanks to all in advance
Dag
Forums: