SNMP agents are responsible for answering SNMP requests from network management stations. Examples of management requests are GET, GETNEXT, and SET, performed on the MIB objects.
A subagent extends the set of MIB objects provided by the SNMP agent. With the subagent, you define MIB variables useful in your own environment and register them with the SNMP agent.
When the agent receives a request for a MIB variable, it passes the request to the subagent. The subagent then returns a response to the agent. The agent creates an SNMP response packet and sends the response to the remote network management station that initiated the request. The existence of the subagent is transparent to the network management station.
To allow the subagents to perform these functions, the agent provides for two types of subagent connections:
For the TCP connections, the agent binds to an arbitrarily chosen TCP port and listens for connection requests. A well-known port is not used. Every invocation of the SNMP agent could potentially use a different TCP port.
A subagent of the SNMP agent determines the port number by sending a GET request for an MIB variable, which represents the value of the TCP port. The subagent is not required to create and parse SNMP packets because the DPI API has a library routine query_DPI_port(). After the subagent obtains the value of the DPI TCP port, it should make a TCP connection to the appropriate port. After a successful connect(), the subagent registers the set of variables it supports with the SNMP agent. When all variable classes are registered, the subagent waits for requests from the SNMP agent.
The query_DPI_port() function is implicitly executed by the DPIconnect_to_agent_TCP() function. The DPI subagent programmer would normally use the DPIconnect_to_agent_TCP() function to connect to the agent, so it does not need to obtain the value of the DPI TCP port.
For a SHM connection, the subagent can use the DPIconnect_to_agent_SHM() function.