Hi Jeff, SReddy, Pando, Tamil, Kishore, Omegamark and others,
I have never implemented OPS environment (though I have read the documentation a little bit)
We are planning to implement it with Oracle 8i (not sure 8.1.5 or 8.1.6) on Sun Solaris
We are evaluating two options for load balancing --
1. Two instances running (on separate nodes) and accessing two different sets of datafiles on two different disk arrays.
The two instances are doing the load balancing and Both the datafile set updating each other as and when there is a change in data.
2. Two instances running (on separate nodes) and accessing the same set of datafiles on one disk array. This will involve the locking issues...what happens when one instance is updating the record that second instance is waiting to read (or update)
I need you opinion about how to implement these types of configuration (what are the pros and cons).
Which is the better way ?
Where can I find some good material on this to read ?
I need to let my boss know today (I am glad he didn't say *yesterday*) which way we are going and why.
I, few months back implemented the fail-over system using Oracle, Veritas and NetApp. This is not anyway OPS but has some similarities in the implementation.
(few years back involved in the OPS implementation, couln't recollect anything now)
I don't think your first option is applicable with OPS. OPS shares the same database (and so the datafiles). You may be able to spread the files across the filesystem, but both instances of the OPS will be accessing them.
1 is not an option of OPS, but Standby or like IBM HACMP failover configuration which we have in our shop.
2 is the right option and by definition it is OPS. Multi nodes(instances) accessing one set of datafiles.
I do have little knowledge on it as we planned to implement it, but we never did. We bought even box to implement OPS and again Management changed their mind and switched over to HACMP to minimise maintenance effort of OPS.