Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Autoincrement in Oracle 8.1.7
By bad app design I am not talking about sequences vs auto increment. I
mean in a very general way. Usually there is a bad application design that
far outweighs the sequence or autoincriment. I once worked in an
environment that was interpreted. The developers would say, "But if we
could only do it in C then it would be fast, interpreted is slow." Without
fail I could walk into any of their shops and decrease the calculation time
by 10x minimum (it was a customized defined benefit system for pension
plans) by making minimum code changes (about an hours worth of work) due to
bad application design. Yes, in C it would have been faster, but I was able
to make some simple changes that really made the whole thing close to C
calculation speed. In this instance the database access (the driver part
was in C and highly optimized) part of the calculation was by far the
slowest part and if you went and read a part of a table 120 times then that
was slower than doing it once 1.
Also I wouldn't use ODBC, too slow, I would use native drivers for whatever
application I was writing.
Jim
"Serge Rielau" <srielau_at_ca.ibm.com> wrote in message
news:3CCD3C63.23EB99F9_at_ca.ibm.com...
> I agree that in a usual application it's not likely to matter that much,
but for
> short and mostly trivial OLTP I bet it's measurable. Your single row
insert
> through ODBC kinda logic the trigger can significantly increase the
pathlength
> even cross compiled to C.
> Also keep data load in mind.....
> Bad application maybe. But changing that old Sybase/MS application is
expensive
> :-)
> As always the backend has dictated the app implementation , so in a
migration
> scenario you won't have any other before trigger logic (because SQL Server
> doesn't have them).
> I think identity and sequences both have their place in the world.
> Identity already is standardized and sequences are in the works.
> But I digress...
>
> Cheers
> Serge
> --
> Serge Rielau
> DB2 UDB SQL Compiler Development
> IBM Software Lab, Canada
>
>
Received on Mon Apr 29 2002 - 09:08:46 CDT