Re: tough choices

I'm reluctantly re-entering thid thread for a guest-appearance....

I s'pose the truth, as so often, is somewhere in the middle. DB2 UDB for LUW's UNION ALL capabilities are bigger than those of Oracle. UNION ALL views support the full range of SQL (update, delete, native insert (without instead of triggers), select) including branch elimination, theorem proving and so on.
DB2 UDB for LUW even collapses a UNION ALL view into a single "parameterized table" if possible, it pushes UPDATE and DELETE through UNION ALL. At this point UNION ALL and range-partitioning effectively move very close together in those DML respects. Obviously these enhancements are influenced by the fact that no table-range partitioning is available in the same way that Oracle reduced investment in UNION ALL technology because of table range partitioning is available.

Nonetheless I see having sophisticated UNION ALL technology as important in the broader context, especially in the context of federated databases. range partitioning across DBMS has to use an SQL solution. After the good experience with TPC-C I feel that, with respect to raw performance, table range-partitioning in DB2 Vx will be hard pressed to beat UNION ALL in DB2 Vx.

I think the bottom line is that DB2 "free" UNION ALL has more to offer than Oracle "free" UNION ALL.
W.r.t. partitioning being more than just range partitioning. Granted, the rides at your favorite theme park are also all free. Pitty if you don't use them all after paying the entrance fee. I sure don't go to themepark because I want to go for a swim and I recall a thread in the Oracle newsgroup where soemone tried hard to use UNION ALL to avoid the cost of the partitioning feature.

Cheers
Serge

-- 
Serge Rielau
DB2 SQL Compiler Development
IBM Toronto Lab