The main factor in deciding to split a topology is the number of managed ports or interfaces per domain and the polling interval and not the number of devices.
The secondary factors influencing the environment setup are :
The solutions that are being positioned, such as IP Manager, Service Assurance Manager, Server Manager, ACM, or VoIP.
The types of devices to be managed, for example, routers, switches, layer 3 switches, load balancers, servers, or virtual machines, and the device vendors in the environment, for example Juniper, Cisco, Dell, IBM, or VMware.
The management requirements such as, operational monitoring, performance, or response times.
The technologies used, such as IP L2/L3, MPLS L3 VPNs, L2 VPNs, or SANs.
The networks that connect the managed devices, such as local, global, cities, T3 links, or metro Ethernet.
The end users and end user locations, such as 24x7 monitoring, follow-the-sun, one NOC, multiple-NOCs, or managed service provider.
To set up the environment calculate the percentage of routers and switches, and then the number of managed ports and interfaces. Refer to the IP and SAM deployment guides to choose the tiers and verify the hardware needed and platform desired. Allot an extra 30 percent for the growth of the network.
For example, if you have 4000 routers with 10 interfaces to be managed, it will generate 40,000 managed interfaces. You will need 2 CPUs and 4GB RAM per domain manager and no more than 20,000 managed ports and interfaces per domain manager, which means you will have at least 2 IP domain managers and hardware that can support up to 4 CPUs and 8GB RAM, which does not include SAM, OI or adapters. If you have more domain managers, then more CPUs and memory are needed.
If the environment is highly meshed, then you will need intelligent topology splitting. With only 4K devices, you do not need the topology split server. It is recommended to have a split based on geography or IP networks for a long term maintenance.