Discovery architecture

Use this information to understand how the components of the discovery process work together to perform a full discovery of the network.

During a full discovery, the Discovery Engine, ncp_disco, detects the existence of devices on the network and queries the devices for inventory and connectivity information. This information is subsequently processed or 'stitched' together to generate a connectivity or topology model.

After a full discovery, the system classifies devices based on a predefined Active Object Class (AOC)1 hierarchy. The network topology is stored in the topology database and can be visualized by network operators as topology maps, customized to show specific devices or specific device groupings such as subnets and VLANs.

The following figure shows how the components of the discovery process work together to perform a full discovery of the network.

Figure 1. Full discovery data flow
Discovery dataflow
 1  Seed devices are discovered
Based on discovery configuration seed settings, the Discovery engine, ncp_disco, sends finders onto the network to find seed devices. The finders discover the existence of devices but do not retrieve connectivity information.
 2  Agents are invoked to identify devices details, connected devices, and device type information
As devices are found, discovery agents are invoked to retrieve details of discovered devices, device connectivity information, and information specific to device types. There are multiple discovery agents to support the wide variety of network devices. Discovery agents interrogate network devices for information using methods such as ICMP, SNMP, SSH, and TELNET.
 3  Connected devices are identified and device details and device type information are retrieved
Discovery agents do not have direct interaction with the network, but instead retrieve information from network devices using the Helper Server. The Helper Server manages the helpers and stores the information that is retrieved from the network. The helpers retrieve information from the network on behalf of the discovery agents. Helpers also translate agent queries into the appropriate network protocol and make requests to the devices.
 4  Connected devices are discovered
Connected IP addresses identified by the discovery agents are fed back into the finders, which discover the existence of these connected devices. Discovery agents are then invoked for these connected devices (step  2 ) and the feedback process repeats itself until the discovery encounters the boundary delineated by the discovery scope settings.
Note: The Discovery communicates with the Active Object Class (AOC) manager, ncp_class, to classify all the devices in the topology based on vendor, type, and model family. It uses the sysObjectId value held in the device MIB and assigns the device a particular classification based on logic held within active object class files.
 5  Network topology is 'stitched' together
Once all devices and device connectivity has been discovered, discovery processing stitchers are invoked. These stitchers 'stitch' together the data gathered by the agents to generate a connectivity or topology model.
 6  Discovered topology is processed by Topology manager, ncp_model
The discovery sends the topology model to Topology manager, ncp_model. The Topology manager processes the discovered topology. For example, it processes linger time for devices and makes appropriate modifications. The topology is also stored in ncp_model.
 7  Network topology is stored
The Topology manager, ncp_model, sends the topology to the Topology database, NCIM. The topology data in NCIM can be queried using SQL.
1 An AOC classifies devices based on vendor, type, and model family.