Re: Distributed transactions (two-phase commint) without DB links
- From: Alexander Gorbachev <gorbyx@xxxxxxxxx>
- To: Martic Zoran <zoran_martic@xxxxxxxxx>
- Date: Mon, 14 Mar 2005 23:57:36 +0100
> > Am I incorrect assuming that DB links are done in
> > principle via the
> > same mechanism?
> Yes, just you do not need to play with API's. You are
> doing your transaction without knowing where, how and
> just commiting at the end.
Well, in our case this IS exactly the problem. Developer has to know
where he does what.
Otherwise:
1) after certain level of abstraction developers stop (or forget) thinking.
2) solutions most of the time slow, don't scale, unreliable (or is it just me?)
> I agree. I hope that sometimes is not happening every
> day, so may be normal to loose something.
> I did not get some books from Amazon last week. They
> will pay for new ones who cares. If you pay less your
> phone or electricity bill, who cares, they will save
> millions on HW instead of some money on our bills.
What about delayed flight? What about 100?
> Do you know how expensive are XA API insert calls in
> regards to normal API calls?
> If you need twice more CPU's to cover that then it is
> going to be more awful, just to have the perfect
> application :)
If XA API calls are so expensive - sure it's not the way to go. In
fact we are still using Tuxedo (6.5) and moving away from it. I
thought if we can use from the client the same mechanism as db links
but under our control from OCI than it would be good alternative.
On the other hand, when on one doing it or supporting it or it's just
too messy - than it's a dead end.
That's the reason I posted here so thanks to all replied!
--
Best regards,
Alex Gorbachev
--
//www.freelists.org/webpage/oracle-l
Other related posts:
