Fallback persistence is used when the primary persistence method fails. Source Address Affinity is a Layer 4 persistence method and is fully supported as a fallback option for most virtual server types.

By default, BIG-IP load balancing algorithms (such as Round Robin) distribute connections evenly across all available pool members. However, persistence profiles override normal load balancing decisions by forcing subsequent connections from a client to be sent to the same pool member.

According to the BIG-IP Administration: Data Plane Configuration documentation:

Persistence creates a client-to-server mapping that is honored before load balancing algorithms are applied.

When persistence is enabled, BIG-IP may repeatedly select the same pool member even if others are available.

This directly alters load balancing behavior.

Why the other options are incorrect:

A. Adding a OneConnect profileOneConnect optimizes server-side TCP connections but does not change which pool member is selected.

B. Enabling SNAT automapSNAT affects source address translation, not pool member selection.

C. Enabling a fallback host in the HTTP profileA fallback host is only used when no pool members are available.

Correct Resolution:

Adding a persistence profile alters load balancing behavior by maintaining client affinity to a specific pool member.

The two load balancing methods that consider all connections between the BIG-IP and each backend node — not just connections to a specific pool member — are Least Connections (node) and Weighted Least Connections (node) .

The critical distinction here lies in the node-level scope of evaluation. A node represents the backend server ' s IP address, regardless of how many services or ports it may be serving. Therefore:

Least Connections (node) directs new connections to the node with the fewest total active connections across all services on that server, providing a holistic connection-count perspective.

Weighted Least Connections (node) operates identically but factors in an administrator-defined ratio weight, allowing servers with greater capacity to proportionally absorb more connections while still evaluating total node-level connection counts.

By contrast:

Ratio (member) distributes traffic based on a static weight ratio and does not dynamically evaluate current connection counts.

Round Robin distributes traffic sequentially in rotation, completely ignoring current connection states on any node or member.

The node-based methods are particularly valuable in environments where a single backend server hosts multiple pool members across different ports, ensuring the server ' s overall load — not just per-service load — governs balancing decisions.