Source Address Affinity

There are a few things to consider here:

1. First, load balancing and persistence, while dependent upon each other, are different things. A (new) load balancing decision is only made in the absence of persistence information. A persistence profile is then used to store a previously-selected decision for re-use across multiple TCP sessions.

    

2. In the absence of a persistence profile, a new load balancing decision will be made upon every new TCP session. So if you remove the persistence profile, you'll have persistence to a single server for the life of that (potentially discreet) TCP session, but if that TCP session closes and the user initiates a new request, a new server may be selected. Depending on the application, this is usually a bad thing.

    

3. Source address affinity is really only useful in environments where the IP space is controlled. Across the Internet, for example, there's no guarantee that traffic won't be NATted, and/or that multiple clients can appear to use the same source address, or worse, a client's IP address can actually change. If it's at all possible, some of the more robust persistence profiles should be used instead. For HTTP traffic, the cookie persistence profile is usually best. For other protocols, you may need to use an iRule to examine unique client characteristics.