How to persist monitor data or how to records server downtime over a period

The typical polling cycle for cacti is 5 minutes, so if nodes are bouncing in between polls that information might get lost. Nagios would be ideal in this case, as it can perform passive service checks. If the LTM enables snmp traps for host down/up messages, then the traps can be sent to a Nagios server which will mark the service (pool member) as down/up, and will track availability metrics.