Re: 9iAS Cluster Distance Limitation

Comments imbedded inline below
with this single all encompassing comment. Not all "clusters" utilize the same technology.

A DB cluster is NOT the same as an OC4J cluster; which may be different from a cluster of infrastructure nodes.

Rhetorical question -
What happens when a/the infrastructure host goes down?

oidba wrote:

> We will be deploying an application using oracle 9iAS and Oracle 9i on
> a Unix platform (not known yet). The application will be deployed in
> 2 data centers that are about 110-130 miles apart. The proposed
> architecture is as follows:
>
> Regional database X 2 (one at each data center)
> 9iAS software cluster X 2 (one at each data center)
> 9iAS infrastructure X 2 (one for each 9iAS cluster)

How big a pipe will exist between the sites? Redundant pipes?
What happens when the network between the sites "breaks"?

>
> There will be cross regional transactions, so one 9iAS cluster will
> access
> the other region's database. The application is one (the same) for
> both regions. So it has to be deployed twice with every code change.
>
> 1) Are there any benefits of having 2 different 9iAS clusters and
> infrastructures?

Yes, trade-offs.

>
> 2) Instead, is it possible to have a single 9iAS cluster running on
> dispersed machines at both centers and managed by a single
> infrastructure that will reside at a single data center even though
> the distance is about 130 miles.

Yes, but see questions above.

>
> The existing architecture calls for each 9iAS cluster to be running on
> 3 servers with software clustering at each data center and the
> infrastructure to be software clustered on 2 additional machines at
> each data center.
>

IMO, you can not really "cluster" web servers or forms servers, but you can configure the environment so that automagical failover can occur from a users perspective. Received on Mon Oct 06 2003 - 15:25:32 CDT