I didn't play with parallel execution in 10g yet, so I have to ask for some consideration here :-) but I guess anything *not* revealed by the // execution plan is implemented using IPC between the QC/consumer/producer processes ... additions/corrections most welcome, Lex. --------------------------------------------------------------------- Jonathan Lewis Seminar http://www.naturaljoin.nl/events/seminars.html --------------------------------------------------------------------- So how does each slave determine the part of the data or the sub-function that it needs handle? -----Original Message----- Waleed, no, Oracle is *not* hiding things here -- as Christian says, the slave processes *do* share the same cursor in 10g. this is a great new feature, especially for Oracle development itself ;-) because from now on they don't need to maintain the "Slave SQL generator" code anymore with every new release of the kernel, which became more and more tedious and time consuming. -----Original Message----- Oracle probably just decided to hide this layer!! -----Original Message----- There's no workaround because it simply run like this in 10g, i.e. coordinator and slaves executes the same SQL statement (you probably noticed the new operations in the execution plan as well...).
BEGIN:VCARD VERSION:2.1 N:de Haan;Lex FN:Lex de Haan ORG:Natural Join B.V. TEL;WORK;VOICE:+31.30.2515022 TEL;HOME;VOICE:+31.30.2518795 TEL;CELL;VOICE:+31.62.2955714 TEL;WORK;FAX:+31.30.2523366 ADR;WORK:;;Pieter Breughelstraat 10;Utrecht;;3583 SK;Netherlands LABEL;WORK;ENCODING=QUOTED-PRINTABLE:Pieter Breughelstraat 10=0D=0AUtrecht 3583 SK=0D=0ANetherlands URL;WORK:http://www.naturaljoin.nl EMAIL;PREF;INTERNET:lex.de.haan@xxxxxxxxxxxxxx REV:20040224T160439Z END:VCARD