As far as I can see the polling hits my entire network with pings to see if my client units are up. Is there a way to randomize the polling so the individuaal units on my network are pinged randomly.
Thanks
I recently helped a Dude developer with a similar issue, and I have done exactly what you request in order to stabilize network traffic coming from The Dude server. I came up with 2 methods, and one being a somewhat different solution, but I think it is relevant to your request.
I have always felt that if things were left at default, and have evidence of this by watching task manager network traffic spiking, that based upon topology, number of nodes, latency and probably other metrics, there will be a very consistent and regular ‘ping flood’ coming from the server. I started to mitigate this by randomizing all the polling parameters at least by map, that smoothed the network traffic significantly. That still might not be enough based upon volume of devices. Breaking maps up into smaller maps, then applying this technique helps more. Again, topology must be considered.
I then came up with another interesting technique based upon a help request. Why not identify and classify devices based upon importance, and/or criticality? If a backhaul or backbone feed is down, it deserves a higher priority. Of course, each device can have its own polling parameters set, but that does incur extra administrative overhead. So, I looked at the ping probe, and realized that multiple categories of the ping probe can be created based upon retries, intervals, etc. With function coding, you could even go further. I think you could smooth out, and actually improve performance, possibly reduce false positives/negatives by having a ‘pinghipri’, ‘pinglowpri’, ‘pingforthesakeofpinging’ set of ICMP probes that would naturally randomize the ping polling.
I dont know if this addresses your specific network issue, but essentially, polling parameters are configurable, and with due consideration, IMHO having different ‘ping’ probe metrics could be a way to get to randomization via a different route. This also creates a more scalable, and easier to configure polling interval that is set just by the use of the appropriate ping probe for a device. Appropriate testing should be incorporated to verify accurate results.
Thinking about this even more, I would guess that utilization of distributed Dude agents could have a similar effect. In some cases it might not be practical or even possible, but just another feature of the Dude that can be leveraged.
Would love to get any feedback on this. It is not the first request for this capability I have received. It served its purpose on another Dude implementation, but need more input from large networks monitored with the Dude. TIA.