Other gotchas to watch out for with synchronous replication:
· Special attention is required to capacity planning for replication if you are running in an HA-pair configuration. You will want to spread those replication relationships equally between both nodes and make sure the number of concurrent relationship in the HA-pair does not exceed 16. If not, when the HA-pair performs a CFO (failover) a certain number of replication relationships will fail. The number of failures will depend on how many relationships above 16 concurrent limit you are trying to run on the surviving node. There is no way to predict which relationships will fail in this situation as it depends on a number of factors.
· Synchronous and asynchronous replication support different concurrent replication operation values. The values are listed in the ONTAP Online Backup and Data Protection Admin Guide. A standard rule of thumb is 1 synchronous relationship = 4 asynchronous relationships (in terms of resources consumed). Also remember synchronous is always consuming system resources whereas asynchronous consume resources when the snapmirror schedule initiates. So, for example, you could have more asynchronous relationships defined than the maximum number of concurrent asynchronous operations supported by the hardware platform as long as all the replications do not occur at the same time (hence a concurrent limitation). You can make this a simple or a complicated as you like (hint : simpler is often better unless you like sleeping only a few hours a day)
