InstantForum.NET v4.1.4Teradata Forumshttp://www.teradata.com/teradataforum/info@teradata.comSat, 21 Nov 2009 22:25:23 GMT20http://www.teradata.com/teradataforum/Topic15653-9-1.aspxYour understanding is still correct for non-PPI tables. With PPI, the row identifier is effectively extended by 16 bits to include the partition identifier.Mon, 08 Jun 2009 11:14:26 GMTJim Chapmanhttp://www.teradata.com/teradataforum/Topic15653-9-1.aspxHi Thanks for the info.........my understanding till now was that....all rows in an AMP will be identified by 32bit hash value+32 bit uniqueness vaue.....Mon, 08 Jun 2009 04:36:42 GMTsreehari Raghavanhttp://www.teradata.com/teradataforum/Topic15653-9-1.aspxKind of wizard, but this is the thing which makes Teradata data distribution so optimal...123 Value will be on same AMP but in different data block(mostly different cylinder also)...Same AMP thing will make it useful for join purpose i.e when these 2 tables will be joined on this column no data is needed to be redistributed as both are on same AMP.....can be processed locally...Whenever a new row is inserted a 3 part message is generated Table ID+row hash+PI value...so a row will definitely go to his companion only, by using Table ID....For more you can Teradata manuals -Physical Implementation.......Regards,PrakharThu, 04 Jun 2009 09:56:43 GMTprakharhttp://www.teradata.com/teradataforum/Topic15653-9-1.aspxif 123 is a value of an unique primary index column of a particular table A,it will generate a particular hash value via hashing algorithm...now suppose a second Table B also has 123 in its unique primary index column......this will also generate the same hash value and will be there in the same AMP.in this case how the AMP will know that which row corresponds to which table.ThanksSreehariThu, 04 Jun 2009 05:56:14 GMTsreehari Raghavan