As you stated you have a few choices, we used one SVM based on the business unit accessing the data. The reason why most people are using different SVMs is to limit access to different business units. We used different VLANS for each SVM to make sure the data is secure. We put the log and temp in a different volumes so we could control the data in the snapshots. If temp and logs are in the same volume the snapshot usage will be very high, which impacts performance. We had each DB in it's own volume. I beleive with RAC you have evertyhing you need to restore a DB so there shouldn't be any need to snapshot the temp/log volume.
I agree with you. The problem is that for this customer we will have to plan with a big accuracy the SVM needed because we're experiencing a lot of issues with DR for test purposes.
We've just started to experiment SVM DR in 8.3.1RC1 but the usage it offer does not permit to replicate the behaviour that vfiler create -d offered us...when we had the choice to create a vfiler starting from a cloned flexvol at destination site maintaining up & running the source that continued to replicate the data.
Dear Eugene, I know fastlane offer, I know courses, I've reached too. But, what we say here in Italy "from say to doing there's a sea between". As experienced netapp ps I prefer to exchange opinions with colleagues here and to read and study the good tr that netapp offers. And last, let me say, you're free to post but I don't like this form of advertising here. Regards