2 Replies Latest reply on Jul 16, 2013 7:49 PM by RichardLetts

    More polling engines or support more element per server

    ctopaloglu

      Our infrastructure is getting bigger everyday. In every 6 months I am being forced to buy a new polling engine license. Because single polling engine (server) supports maximum 12.000 elements (8.000 recommended). This must be a joke! The competitor product supports 1.000.000 (One million) interface(element) per server! So, better scalability please!

       

      I have created an idea for this. Please support if you like.

       

      http://thwack.solarwinds.com/ideas/2537

        • Re: More polling engines or support more element per server
          Leon Adato

          Mustafa, I'm not sure what tools you are using or what hardware/OS platform they are running on, but let me offer my experience:

           

          1) I just shifted from IBM Tivoli to SolarWinds. There was NO single server in the Tivoli infrastructure that could support SNMP polling of 12,000 elements of anything at anything resembling reasonable time frame (meaning: Your one Solarwinds server can poll 100,000 elements as long as you only poll things once per day. Once we get down to the default every-5-minute level, we're done.

           

          In fact, a rule of thumb that I've observed over the last 11 years of building and maintaining monitoring systems (Tivoli, BMC Patrol, Nagios, Zenoss, SiteScope, etc) is that any single server performing a "polling" type action typically chokes at 10,000 elements. It has more to do with the number of open file handles which is built into the OS (especially windows, but Linux isn't infinite either) than anything else.

           

          Interestingly enough I've thrown boatloads of CPU and RAM at a poller, and had no appreciable improvement once I hit a certain threshold (around 10,000 or so). Meaning that once your server tops out in terms of elements, throwing more hardware at it is not going to make the problem go away.

           

          2) Solarwinds *CAN* handle more than 12,000 elements. You just need another polling engine. That's why they sell additional pollers. NOW.. you wanna talk about the $COST$ of those pollers, that's another story.

           

          In point of fact, one Solarwinds "installation" (primary poller plus as many additional pollers as you wanna buy) can grow to about 110,000 elements. After that you can probably keep going but support isn't going to be nice to you on the phone.

           

          3) I'm curious who you are naming as the competitor. We just did an agentless RFP. UpTime, HP, and Zenoss came in around the $1million mark. OpenNMS and  Nimsoft never responded. Nagios, Shinken, Cacti were all non-starters.

           

          SolarWinds? Hit 89.9% of our RFP and came around 10% of the other bids.

           

          Explain to me who competes with that.

           

          Yeah, I'm a fanboy. What of it?

            • Re: More polling engines or support more element per server
              RichardLetts

              In fact, a rule of thumb that I've observed over the last 11 years of building and maintaining monitoring systems (Tivoli, BMC Patrol, Nagios, Zenoss, SiteScope, etc) is that any single server performing a "polling" type action typically chokes at 10,000 elements. It has more to do with the number of open file handles which is built into the OS (especially windows, but Linux isn't infinite either) than anything else.


              this is only the case for bad design: you can poll the entire Internet of devices using only one filehandle. I've written such code in the past (1998/1999), it just requires a little more thought about how to do large-scale SNMP.


              Richard