Used nagios to monitor about 300+ machines
1) easy installation and support — there are packages in many distributions, adjusted simply
2) extensibility — a lot of plugins, including network equipment, you can write your own, it will be enough to write a shell script that writes the data in a particular format.
3) the capacity for 300+ vehicles, the server lacked 256 MB, while it has worked OCS Inventory
4) reliability — like a brick, very reliable.
5) data visualization — there is beauty for everybody.
6) distributed monitoring — I do not quite understand what I mean. If that's what I was thinking, there is NRPE, which executes plugins remotely on the host and gives the response of the plug-in server monitoring. If you mean using multiple monitoring servers in tandem, it is not certain that the company in 50-70 cars need it.
7) escalation of incidents — not ready to say anything specific, no idea how you have the company with this situation, what is the hierarchy of incident management. Based on the size of the company, one technical specialist and a couple management.
8) a wide range of alert — it is possible to divide users into groups and specify which group to send alerts, normally there will be three service status or parameter are OK, WARNING, CRITICAL. There are still status UNKNOWN, arising in the event of incorrect operation of the plugin.
9) price — free
Plus there is the ability to execute scripts remotely in the event of a critical situation. For example, to distort the service.
In addition, you can use the coordinate drawing on the map. Define coordinates for all machines in the diagram and see where it is. However, you will have to attend to the task coordinates.
Cacti is more suitable for monitoring via SNMP, consequently, they monitor the network hardware, I personally used to do plugins for nagios'and.
In General, from my experience I suggest nagios.
cacti is more suitable for monitoring network equipment — it charting very comfortable.
zabbix — rather like the light version of Nagisa.
nagios is a very agile monster. but the interface is so beautiful not that much scary.
opsview — add-on nagios'ohms. very beautiful.
on the basis of opsview easily created distributed network monitoring. You can create clusters.
About the second question.
HP OpenView did not feel, but the Tivoli approach to your goals better than all of the above, but there is only one nuance — the price is clearly not for a small company.
Myself suggest Cacti(Nagios) + Munin