|
Posted by --CELKO-- on 02/07/06 17:35
My point is a that slow code is usually the result of bad DDL, whcih in
turn leads to complex DML. You do not start by adding indexes and
hints. You fix the leak instead of mopping the floor, over and over.
But you remark is interesting, A friend of mine who writes database
engines for a major vendor observed that SQL Server people do not trust
their optimizer like Ingres, Postgres, DB2 people do. That is a bad
thing to say about the product.
The classic one was a test done by Fabian Pascal in DATABASE
PROGRAMMING & DESIGN years ago. It should still be in the CMP archives.
He wrote the same query seven different ways and ran them on various
SQL products on the same desktop hardware.
Ingres knew they were logically identical and produces the fastest
results of any
of the products. Oracle produced different execution plans for each
query and the slowest running one had to be shut down after grinding
for hours instead of seconds. The other products were all over the
place. But by now, everyone who is stil in business has improved since
those days.
Navigation:
[Reply to this message]
|