Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: oracle grid

Re: oracle grid

From: Joel Garry <joel-garry_at_home.com>
Date: 29 Sep 2004 16:20:08 -0700
Message-ID: <91884734.0409291520.697e78a6@posting.google.com>


hjr_at_dizwell.com (Howard J. Rogers) wrote in message news:<14a1f766.0409231440.5a251f5d_at_posting.google.com>...
> joel-garry_at_home.com (Joel Garry) wrote in message
>
> > > I'm going to simultaneously agree and disagree with you Daniel!
> > >
> > > You *can't* have a Grid without a RAC (so Noons is right). For the
> > > simple reason that, without RAC, as an end-user, I would know what
> > > server I was connected to (ie, I would *have* to know), whereas the
> > > essence of Grid-ness is neither to know nor care. And for that, I need
> > > a choice of instances running on a plethora of servers.
> >
> > End users don't *have* to know. I have middleware that hides it from
> > them. So, for example, if Mr. DSS-guy has to try every whichwaything
> > with his data, I can put a replicant on his pc (or if he complains
> > about performance, let him buy another pc/server) and let that take
> > the hit, rather than the oltp system. Perfectly doable with O8
> > vintage middleware.
>
> As I said, there are merely degrees of gridness; a spectrum. It's not
> black and white. What you've described is at the low-end of gridness,
> and yes it was perfectly do-able in 8.0+.
>
> But clearly, your description is not very griddy. As an end-user, it
> would seem that I have to know whether I should attach to a local
> replica of the data, or whether I can use the main infrastructure. As
> an administrator, it would appear I am supposed to know whether this
> guy's queries are clobbering the OLTP users, or whether I need to do
> something about it.
>
> In a "true grid" paradigm, neither of those bits of information would
> be needed, because the backend server would just reconfigure itself to
> make sure the DSS stuff runs without affecting the OLTP stuff. And if
> that involved powering up another instance to dedicate to the task, so
> be it. There would never be a question of having to replicate to a
> desktop, because the backend would sort itself out so that it was
> always up to the job. Perhaps another "grid technology" (formerly
> known as Resource Manager) would kick in to sort out the competing
> demands... who knows? Point is, it gets sorted, without my having to
> poke around with init.ora parameters etc etc etc.

Ah, I think you've hit on something that has been bugging me about all this: I've never seen any place (other than dot-bombs) that has enough resources, much less enough to allow autonomic reconfiguration.  YMMV.
>
> Similarly, your o8 vintage middleware does a reasonable job of hiding
> the complexities from users -but in the very griddy 10g world, we have
> clusters of 10g Application Server, so that even on the middle layer,
> we have no idea which particular middle server you've connected to.
> And those multiple application servers load-balance and re-configure
> themselves so they can always cope with the load.
>
> Do I believe any of this stuff is for real? Not much. I'm merely
> playing devils's advocate and trying to interpret the Oracle marketing
> hype (for myself as much as for anyone else). And truly, there's not
> much fundamentally new about most of the technologies needed to make
> it happen -it's just that 10g has brought them all together and
> improved them and co-ordinated them to the point that bits of the
> dream can become a form of reality. If it were truly do-able, I'd sign
> up for it. And 10g certainly gets us some of the way there. It
> certainly doesn't get us *all* the way there, but on the other hand I
> think that saying most of it was doable with version 8i is rather
> missing the point... about how *well* it was doable, or how *easily*.

Yeah, depends on how easily easily is. The real scary part boils down to the old UI v Command Line argument, which is, in making it easy, the complexities are hidden in some translation layer. And if they are hidden, they are no longer easy - at some unpredictable point of complexity. You don't know when or how it will break, but it certainly will.

<rant>
dammit, I used to be able to pick up a phone and be fairly sure it was going to work, and keep working until I was done with it! </rant>

jg

--
@home.com is bogus.
cdos stats:  http://groups-beta.google.com/group/comp.databases.oracle.server/about
Received on Wed Sep 29 2004 - 18:20:08 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US