How is the "failover" method implemented in the General Purpose model? Is it Windows Clustering with shared disks so if the primary SQL Server is offline, one of the other SQL Servers will "mount"/take ownership of the disks? Or is it something special built for Azure?
Azure SQL only allows one replica to serve both read and write workloads so load balancing isn’t really a concept here. You can redirect read-only workloads to a secondary replica. Does this answer your question?
The thing is, the replica is still in the same region right? So you're probably better off using GP and creating your own Auto-failover group with a second MI in a second region? That way you've got multi-region HA, which is better than what Business Crit offers.
It just depends on your use case. If your main goal is offloading read-only workloads then using the included replica of the BC tier meets that purpose. The replica is there for high availability purposes and to ensure Microsoft can maintain its SLA’s. But if you are looking for a solution for disaster recover, then, as you stated, auto-failover groups would be the best solution for that use case.
I am analysing here for my company and converting 40 vCore of general to 16 vCore of Business Critical I will get 4 x more iOPS and pay almost the same price.
Business Critical is a great option for when you need more IO as it uses locally attached storage vs the remote storage in the GP tier. I know you are using Azure SQL MI, but if you have a need for Azure SQL DB, the Hyperscale tier is a great option to look into.
@@DataBar Thanks for the answer... Yes, we are using MI and need to stay there... Have You experienced some conversion like that? Our intention is to scale up the tier but scale down the number of vCores to keep the budget aligned :) thx...
@@DataBar Yes for sure I'm considering this as well... I'm impressed on how the iOPS are different per vCore... Pay the same amount to have 4 x more it's definitely an advantage.
How is the "failover" method implemented in the General Purpose model? Is it Windows Clustering with shared disks so if the primary SQL Server is offline, one of the other SQL Servers will "mount"/take ownership of the disks? Or is it something special built for Azure?
Yes it is very similar to how Windows/SQL clustering work. And the business critical tier failover process is based on AlwaysOn AG groups.
Great explanation to make decisions
Glad it helped!
Since the solution is centered around BC and zone redundancy could you use lb as opposed to traffic manage to reduce cost ?
Azure SQL only allows one replica to serve both read and write workloads so load balancing isn’t really a concept here. You can redirect read-only workloads to a secondary replica. Does this answer your question?
The thing is, the replica is still in the same region right? So you're probably better off using GP and creating your own Auto-failover group with a second MI in a second region? That way you've got multi-region HA, which is better than what Business Crit offers.
It just depends on your use case. If your main goal is offloading read-only workloads then using the included replica of the BC tier meets that purpose. The replica is there for high availability purposes and to ensure Microsoft can maintain its SLA’s. But if you are looking for a solution for disaster recover, then, as you stated, auto-failover groups would be the best solution for that use case.
I am analysing here for my company and converting 40 vCore of general to 16 vCore of Business Critical I will get 4 x more iOPS and pay almost the same price.
Business Critical is a great option for when you need more IO as it uses locally attached storage vs the remote storage in the GP tier. I know you are using Azure SQL MI, but if you have a need for Azure SQL DB, the Hyperscale tier is a great option to look into.
@@DataBar Thanks for the answer... Yes, we are using MI and need to stay there... Have You experienced some conversion like that? Our intention is to scale up the tier but scale down the number of vCores to keep the budget aligned :) thx...
Definitely make sure you take advantage or reservations. That’s the easy way to reduce cost for your workloads.
@@DataBar Yes for sure I'm considering this as well... I'm impressed on how the iOPS are different per vCore... Pay the same amount to have 4 x more it's definitely an advantage.