Re: Options for wide-area database?

Notice that my intent is to deploy a wide-area oracle, DR/BC only comes in as bonus. The main concern is to serve clients across a wide distance with the highest data consistency possible. My real question is: usually a nation-wide db will be deployed with advanced replication, but if RAC can actually maintain the same db over multiple servers (vs replication data that is not up-to-date), can't we spread the distance of RAC servers if those servers can see the 'same disk'?

'Same disk' --> basically using a wide-area SAN or wide-area NAS.

Sorry for my ignorance, but I'm not sure what you mean by the comment below:
> > 4. Duplicate the SAN, or use NAS to save money, and spend your money
> > on network.

By duplicating SAN, that sounds like I have to do advance replication.  Actually is option 4 even a viable solution at all (1000miles!) because the RAC cluster exchange at host level may kill the performance. Anybody's done that before?

Ernest

joel-garry_at_home.com (Joel Garry) wrote in message news:<91884734.0402061620.7419441f_at_posting.google.com>...
> Daniel Morgan <damorgan_at_x.washington.edu> wrote in message news:<1076014857.358088_at_yasure>...
> > Ernest Siu wrote:
> >
> > > What are the pros/cons for the following options in implementing a
> > > wide-area oracle (e.g. 100-1000 miles apart)? In terms of cost,
> > > performance and data consistency?
> > >
> > > 1. Single-site single-server (big!) with remote clients
> > > 2. Single-site multi-servers using RAC with remote clients
> > > 3. Single-site server(s) (option 1 or 2) with advanced replication
> > > onto remote hosts
> > > 4. Multi-site servers with RAC and wide-area SAN
> > >
> > > Any comment/suggestion?
> > > Ernest
> >
> > What are you trying to achieve? What is the cost per minute of downtime?
> > In dollars? Without this information any answer is meaningless.
> >
> > 1. Bad idea unless you don't believe in Murphy's Law. Spend your money
> > on hardware.
> >
> > 2. Bad idea unless you don't believe in Murphy's Law. Spend less money
> > on hardware.
> >
> > 3. Reasonably good but spend your money on network.
> >
> > 4. Duplicate the SAN, or use NAS to save money, and spend your money
> > on network.
> >
> > My approach would be two data centers at least 1000 miles apart. Both
> > with a NAS or SAN and each with two or more RAC nodes. At each site use
> > DataGuard.
>
> I second the comments about the network, after trying to maintain a
> standby where the backed up db is much larger than the pipe.
>
> I disagree about the NAS, although concede I might just have had bad
> luck.
>
> jg
Received on Mon Feb 09 2004 - 12:06:42 CST