It is essential to take care with how one handles information when operating load balancing. During a user’s sessions, one piece of information may be subject to multiple requests. For example, if certain information is stored locally on just one backend server, subsequent requests journeying to different backend servers would by no means be able to locate it. Persistence is when all requests made during one user session are sent consistently to the same backend server. It is a useful technique yet if a single back server goes down, its information becomes inaccessible during the session. Also, any sessions that depend upon that information are subsequently lost.

It is not straight forward as to what particular server one is assigned to. It can be a random assignment or it can be due to a client IP address. However, this method is rather unreliable as the load balancer must recall all the random assignments which inevitably leads to a burden on storage.