I agree with your idea of being able to stop backups to beef up performance when its needed (batch processing etc.). I would like to add another scenario/add-on to this.
I believe we need to be able to declare business hours (or heavy load hours) for the instance. As this is not available, full backups and differentials can run during business hours. We have seen this in action and it reduces the level of confidence in the solution to run at its best consistently. I say this as when building a DB solution regardless of the purpose (OLAP, OLTP, etc.) the one thing you need is consistency, which you can not get if an automated backup process decides to run during business hours.
To confirm this would be fine if the backup process did not effect the performance of the MI, but we have many examples of the negative impacts that full and diff backups have on an MI with heavy workloads. I believe Hyperscale will solve this issue in the future, but if traditional MI/SQL DB's are still planned to be available as a service moving forward, this will need to be addressed.
Hi Ken, I agree the build times are frustrating, but I believe the best way to provide a solid DR plan is to deploy a 2nd instance in another region close to you and configure it as a readable replica in a failover group. I know this is annoying as it essentially doubles the cost for GP instances, but overall it provides quite a nice solution for those instances that the business has an SLA of less then 6 hours etc.
All the best,
3 votesMatt Mellifont shared this idea ·
I also feel this pain, playing around with blueprints to create standard environment templates to deploy and always finding the VPN takes an age to deploy in AU east