Snmp push mechanism

Question

Recently I am stuck into a snmp problem. My client requirement is that I have to push cpu,hard disk and memory data to the Network management system periodically. I have already configured my snmpd.conf file to pull data through snmpwalk command but don't know how to push periodically to nms. Also need to know how to test that the data's are pushed properly or not. Anly help would be great

Answer 1

If the NMS provides any other interfaces than SNMP, for example, 3GPP XML files transferred over FTP, I would strongly recommend using that interface instead.

You have an SNMP Agent running already, and it will respond to get-requests, walk, etc. Normally, an enterprise-grade NMS would have no problem polling an SNMP Agent regularly to collect data such as what you describe. This is a common approach in what the telecom sector defines as Performance Management(PM) according to FCAPS. For a modest amount of counters, fetched at reasonably large intervals, this approach generally works well. Problems with polling time can sometimes occur with too-frequent polling of large amount of data.

From the SNMP Agent, you also have the option to send Trap messages, which are spontaneous asynchronous messages. Normally, traps are only used to notify an NMS about important events on the supervised equipment, such as equipment faults (Fault Management). However, there is technically nothing stopping you from designing a MIB which defines traps sent regularly, containing performance data. Some form of adaptation would probably be needed on the NMS, to receive PM data from SNMP traps, since this is not usually done. If the NMS is not able to do regular polling of counters, it seems unlikely that it would be flexible enough to do this.

If there is a large amount of counters, traps are unsuitable since the size of each message should ideally not exceed the MTU of the network (1500 bytes for Ethernet).