6 Replies Latest reply on Jan 19, 2010 8:19 AM by ecornwell

    Additional load for monitoring access switches

    SimonMorris

      Hello,

      I wanted to ask the opinion of the forum about using NPM to monitor "user facing" Ethernet ports on Access layer switches.

      It would be handy to know if a port hasn't been used for a while, but I'm concerned that monitoring hundreds of additional interfaces is going to add load to the Orion server.

      Bearing in mind that NPM goes to the device once every 10 minutes anyway to get it's status, does it really add a lot of overhead to collect the interface statuses also?

      Do other users collect stats for all interfaces on the network?

      Thanks

      ~sm

        • Re: Additional load for monitoring access switches
          lchance

          MY OPINION

           

          We have the ‘unlimited’ SLX license but you definitely have to balance that out with hardware on Orion and the hardware for SQL. We use two servers – Orion and Orion’s SQL.

           

          At first I was reluctant to monitor every switch port but after Orion offered a ‘unpluggable’ feature for port monitoring it helped make sense for something like you said “…to know if a port hasn't been used for a while”.

           

          But having hundreds of ports can become a maintenance nightmare. We only have 2,800 being monitored and the polling and data collections have been fine with no major issues. But I know some folks have ten times that number and they have to have multiple Orion’s – that is, extra Polling Engines to pick up and distribute the load.

          1 of 1 people found this helpful
          • Re: Additional load for monitoring access switches
            ecornwell

            Hi sm,

            We monitor all our switch ports.  We were using MRTG before moving to NPM and it monitored all the ports and we enjoyed being able to see who is using what kind of traffic.  We've got nearly 20k interfaces that we monitor.  We've thought about slimming it down, especially at some of our remote sites.  It seems that just about the time we do, we run into something that we can use the data to help troubleshoot or prove a case for something.  I'm happy we do it but it puts a strain on the SQL server.

            Hope that helps...
            Eric

            1 of 1 people found this helpful
            • Re: Additional load for monitoring access switches
              seigniory

              From my experience, even when doing a full snmp-walk on a device with 11k+ OIDs (snmp-informant), the device's CPU never deviates from baseline by more than 1-2%.  As long as your poller can handle the additional datapoints, (and SNMP polling is properly implemented on the host to control process usage) I say go for it.