0

Does exists any best practice/guide lines of the maximum network latency (ms) between two synchronous AlwaysOn replica?
I need to give this information to network administrators for configuring the link between the two nodes of the cluster.
I know that syncronization between replicas is influenced by disk speed and other things: The picture in this article explain the process involved.
The AlwaysOn instance will host the sap database (1 TB of data) and I estimate that the workload will be considerable with large transactions.

  • The best practice is 1ms latency. As for the maximum tolerable, that depends entirely on how much performance you're willing to sacrifice for the benefit of synchronous replication. – Tony Hinkle May 20 at 12:00
  • 1
    @TonyHinkle is that documented somewhere, or personal experience? – Josh Darnell May 20 at 13:00
2

There is no maximum for the supported latency for synchronous replicas. You will sometimes opt for a higher-latency deployment in return for a RPO=0 for a broader range of failures.

The most common mission-critical configuration is to have a pair of synchronous replicas on separate racks in the same datacenter, and a third, asynchronous replica in a separate datacenter.

But guidance for your scenario should probably come from SAP, or from the Microsoft/SAP Engineering Alliance. See eg https://techcommunity.microsoft.com/t5/running-sap-applications-on-the/bg-p/SAPApplications/label-name/AlwaysOn

| improve this answer | |

Not the answer you're looking for? Browse other questions tagged or ask your own question.