Re: Where is Oracle’s Grid ?

Mark.Powell_at_eds.com (Mark D Powell) wrote in message news:<2687bb95.0312190850.6f64ee55_at_posting.google.com>...
> Daniel Morgan <damorgan_at_x.washington.edu> wrote in message news:<1071675255.966810_at_yasure>...
> > Niall Litchfield wrote:
> >
> > > "Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message
> > > news:1071649660.660848_at_yasure...
> > >
> > >>Actually not. Having spent the past weekend building RAC nodes and
> > >>working on the differences between RAC and grid I can assure you that
> > >>they are very substantial. There is not only no farce ... there is a
> > >>technological improvement here that is incredible.
> > >>
> > >>The NDA, or perhaps my fear of it, prevent me from going into detail.
> > >>But the differences are very significant: At least as large as OPS to
> > >>RAC and that was a complete rearchitecture.
> > >
> > >
> > > Hmmm. I don't thing that I'd describe RAC as 'a complete rearchitecture' of
> > > OPS. I think it would be better to describe it as a fixed implementation of
> > > OPS. That doesn't mean it isn't a huge improvement, but it certainly wasn't
> > > pull out all the old OPS code and replace it with nice shiny new version 1
> > > RAC code - if it was it sure as hell wouldn't have worked apart from
> > > anything else.
> > >
> > > cheers
> >
> > I'll most respectfully disagree. Think back to OPS and the block
> > pinging. The inability to share blocks between nodes. With RAC's cache
> > fusion the architecture is completely different.
>
> We have run OPS since version 7.0 and on version 8 while working with
> support on OPS specific bugs support referred to our bugs as cache
> fushion bugs. RAC has been in development for a long time no matter
> Oracle's official line to the contrary.
>
> RAC is a fixed OPS but I have been told RAC is not truely that hard to
> break. We have some small RAC systems now. By next January we should
> have a good sized on one on 9.2.0.4 and we will see once we
> departition our application how well it works.
>
> I wonder why Oracle did not choose to send only the change data to the
> instance which holds the dirty block rather than transfer the entire
> block to the last instance that wants to update it. Oracle probably
> has a good reason but just sending row data would decrease
> interconnect traffic.

I would guess that more than one instance would have some horrific consistency issues possible. Much easier to do it hot-backup-redo-log style?

>
> IMHO -- Mark D Powell --

jg

--
@home.com is bogus.
http://www.nytimes.com/2003/12/21/magazine/21IDENTITY.html?pagewanted=all