Remote-query Vs. Manual SNMP polling

We use both methods (remote-query and direct snmpget)

Firstly, on the CMTS we have remote-query setup for interval 10, and this seems to work fine with 5000 modems on a PRE4. The readouts probably lag 10 to 15 mins behind realtime which isnt too bad.

Cisco recommend you start with the default 30, and then wait and monitor the polling responsiveness. If you need the polling to be faster, then drop the interval a bit. If your load is too high then increase the interval. More info here http://www.cisco.com/en/US/docs/ios/cable/command/reference/cbl_06_cable...

Secondly, we have a script that logs into the CMTS every 30 mins and collects all the remote query results (eg show cable modem phy) plus other useful stats, and saves them to a mysql database.

Then we have some web frontend tools that snmpget's live readouts plus also presents historical counts/graphs from MySQL for hit/miss/padj/flap/ussnr/uspwr/dssnr/dspwr/rxpwr etc.

Also, we have other troubleshooting tools that direct SNMP poll modems.

For example one script will retrieve a list of CM IP from our provisioning system for a given city or node, and then asynchronously snmpget some stats from each modem, and then can plot results onto google maps as colored pins. Results might be as simple as up/down status (to let you spot outages), or other stats like power levels or USSNR so you can look at a map and see clusters where there is signal troubles.

Perl Net::SNMP can poll 5000 modems in just a few seconds!