Enhanced load balancing on L2TP physical service interfaces

When a service interface in a service device pool is rebooted, sessions reconnect and new session requests are distributed based on the number of sessions on the available interfaces in the pool. The sessions are assigned to the interface with the fewest sessions. If more than one interface has the minimum number of sessions, then a random selection determines which interface gets the session. In earlier releases, session load balancing is a simple round-robin distribution among the interfaces. Consequently, fewer sessions are assigned to a newly rebooted interface than to the other interfaces. For example, consider a pool with two si interfaces, si-0/0/0 and si-1/0/0. Each has 100 sessions. If si-1/0/0 reboots, it drops all 100 sessions. As the sessions reconnect, they alternate between the two interfaces so that when all sessions have reconnected, si-0/0/0 has 150 sessions and the reconnected si-1/0/0 interface has only 50 sessions. Consider the same pool with the new behavior. As sessions reconnect, si-1/0/0 has fewer sessions (0 to start) than si-0/0/0 (100). Because the interface with the fewest sessions is selected, all sessions are assigned to si-1/0/0 until it reaches the same count as si-0/0/0. For asi interfaces, the interface with the lowest session count is selected from the pool for new or reconnect session requests. When the active si interface in the asi bundle goes down, all the active sessions on that primary interface fail over to the secondary interface.