Mirroring

Mirroring connection information involves sending a copy of the mirrored flows to the other BigIP for processing - it doesn't just send summary or state information, so this means that the network path your mirror traffic flows over needs to be able to handle the total bandwidth of all data being mirrored. If you're mirroring a single virtual server, for example, then the total traffic to and from that server will be mirrored, in order for both boxes to maintain the same connection table and thus ensure a seamless failover. This is only really necessary when your applications can't tolerate a transport layer failure on a long lived connection. For short lived connections such as HTTP requests, there's little value in mirroring the flows.

But if you're only mirroring persistence data, then this reduces your bandwidth requirements significantly, as only the persistence records are mirrored