differences for working of cookie and source address affinity

Pretty straight forward. All affinity mechanisms are designed to provide persistence to a single "chosen" server one an initial load balancing decision has been made. And how that persistence is maintained is a matter of the mechanism itself.

Source Address Persistence and pretty much all of the other affinity methods other, than cookie, maintain a "state table" on the BIG-IP - an in-memory table that tracks session persistence relative to some unique attribute of the client (or server). In the case of source address, the table is keyed on the IP address of the client. Source address persistence is a reasonable affinity method in many case, but will have issues in situations where source addresses may be NATted. All of this persistence methods also put the burden of "state" on the BIG-IP and its memory.

Cookie Persistence is of course only useful in HTTP communications, but unlike the other methods, puts the burden of state management on the client. In this case, all of the information that the BIG-IP would have stored about a persistent connection are now sent directly to the client in the form of an HTTP cookie. The client sends this cookie back in every request, therefore maintaining that affinity.

Here's one a several references on the topic of session persistence: https://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/ltm-concepts-11-3-0/ltm_persist_profiles.html?sr=56617015