Re: Equivilant of Oracle's DB Links in DB2
Larry wrote:
[color=blue]
>
> And for the 300GB TPC-H?
>[/color]
Larry - Are these questions purely rhetorical or do you really not know ?
By eyeball and quick back of the envelope calculation, the best 300Gb
DB2 result has about 19K of scripts in the appendices, the closest 300Gb
Oracle result has about 11K.
[color=blue]
> Not the point, Mark. If you're more than happy to get into that
> discussion, then you should have demonstrated that when you made your
> first posts instead of just inserting your editorial comments.[/color]
I think my first editorial comment was "Seems a little overkill for a
simple DB2 to DB2 link (why do I need a wrapper and a server definition
- won't the nickname suffice ? That's effectively all that is required
in Oracle)." at which stage I would have hoped that a "It's for
performance reasons, dummy" conversation could have started. I mean, I
did ask, right ?
So let me ask again. What value does defining the wrapper and server
have when the target is a known entity ? You know the protocal, you know
the version, you know the dialect and data type support, and you know
the optimizer capabilities of the target. Why externalise these ? Is it,
as I summised, that the greater case (access to unknown entities) has
subsumed the simple case (which, BTW, is a perfectly valid approach), or
is there indeed some performance/ease of use/adherence to CS best
practices value that I am missing ?
Cmon. Throw me a bone. Teach an old product manager a few new tricks. I
love to beat up on our developers and tell them how things _should_ be done.
Larry wrote:
[color=blue]
>
> And for the 300GB TPC-H?
>[/color]
Larry - Are these questions purely rhetorical or do you really not know ?
By eyeball and quick back of the envelope calculation, the best 300Gb
DB2 result has about 19K of scripts in the appendices, the closest 300Gb
Oracle result has about 11K.
[color=blue]
> Not the point, Mark. If you're more than happy to get into that
> discussion, then you should have demonstrated that when you made your
> first posts instead of just inserting your editorial comments.[/color]
I think my first editorial comment was "Seems a little overkill for a
simple DB2 to DB2 link (why do I need a wrapper and a server definition
- won't the nickname suffice ? That's effectively all that is required
in Oracle)." at which stage I would have hoped that a "It's for
performance reasons, dummy" conversation could have started. I mean, I
did ask, right ?
So let me ask again. What value does defining the wrapper and server
have when the target is a known entity ? You know the protocal, you know
the version, you know the dialect and data type support, and you know
the optimizer capabilities of the target. Why externalise these ? Is it,
as I summised, that the greater case (access to unknown entities) has
subsumed the simple case (which, BTW, is a perfectly valid approach), or
is there indeed some performance/ease of use/adherence to CS best
practices value that I am missing ?
Cmon. Throw me a bone. Teach an old product manager a few new tricks. I
love to beat up on our developers and tell them how things _should_ be done.
Comment