This discussion has been locked. The information referenced herein may be inaccurate due to age, software updates, or external references.
You can no longer post new replies to this discussion. If you have a similar question you can start a new discussion in this forum.

Second Poller and WAN Traffic

Currently we have two sites with a 2Mbps pipe in between, we have our main Orion server at one site and a second poller at the other.

Now as I understand it the second poller poles locally then sends its data to the primary server, why is it that when the pipe is running at 60 - 70% the stats and polling info from the second poller shows packet loss and in some cases devices being down even though the poller is polling locally and then sending it’s data to the primary server, this never happens when the WAN is running < 40%?

The local devices when checked are not down and not loosing packets.

Also what sort of impact does a second poller have on a network ie what utilisation would it use on a 2Mbps pipe with 1400 elements could it use 15% - 25% of the available bandwidth?

Thanks


IT Infrastructure Manager
Pilgrim Hospital
UK
  • IIRC, the pollers primarily interact with the SQL database. The way the events and such are laid out, there should be a relatively constant amount of data flow from the poller.

    That said, depending on your stats, you're looking at an average utilization over a period of time.

    Based on your description of the WAN link and the problem, it sounds like you're running on a frame-relay or ATM solution. In my experience, most of those are provisioned with a committed information rate at 50% of the maximum. So if you had a full E1 as a frame, you would have a CIR of 1mb. Anything above the CIR is subject to discardability depending on utilization within the provider's network.

    The measured impact is tough to calculate, but can vary significantly depending on how you have your polling set up. The best way to measure this absolutely is to poll out one NIC and send DB updates out another, but that also assumes your DB and what you're polling are on different subnets.

    I haven't completed my deployment yet, so I can't share my traffic percentage yet. emoticons_sad.png

    // Ian Underwood - Service Management
    // Level 3 Communications
  • Thanks for the info

    What I don't get is before we got the extra poller we were polling across the WAN and that did not have any impact on the utilisation like this has.

    Presumeably if I change the polling interval and stats collection time then the utilisation should go down?

    Thanks

    IT Infrastructure Manager
    Pilgrim Hospital
    UK
  • The polls themselves are pretty simple and straightforward, and will definitely take up less space over the WAN.

    With a poller across the WAN, it needs to talk to the database for several items, and not just to dump polling data. The poller also needs to refresh its device list from somewhere, though I'm not sure what the frequency is.

    Presumably, you could increase the collection times, and it could reduce your network overhead, but the back-end conversation between the poller and DB has the potential to be a lot more chatty. A couple of pings is going to use much less bandwidth than the SQL transaction to write a min/max/average time to the database.

    // Ian Underwood - Service Management
    // Level 3 Communications