Page 2 of 2 FirstFirst 12
Results 11 to 12 of 12
  1. #11
    Untangle Ninja sky-knight's Avatar
    Join Date
    Apr 2008
    Phoenix, AZ


    Tests run with the UVM off tell you what the top speed of the appliance are.

    Tests run with the UVM on tell you what you're going to get after the weight of the UVM. Unless you've configured QoS, this performance loss is felt symmetrically. I can not explain your asymmetric loss from the UVM itself.

    What I can say is CPU is latency, RAM is throughput. Overloads of either due to NIC driver processing WILL NOT SHOW in user land, which means you will never see it in the performance graphs you're looking at.

    There's also this:

    There's a final curve ball... the PCIe bus. Desktops are notorious for not having enough lanes to saturate THREE gigabit NICs. Which is required to get 1 gigabit of through put.

    1 WAN NIC
    1 LAN NIC
    1 SATA Link

    If any of those things can't keep up... Untangle gets slower. Each of them is 1gbit. UVM really leans on that last one, turning it off removes it from the mix. So it may mean you can do 2gbit of traffic on the mainboard but not quite a 3rd. Again, building appliances that can do gigabit filtration is hard.
    Last edited by sky-knight; 06-29-2021 at 08:30 AM.
    Rob Sandling, BS:SWE, MCP
    Phone: 866-794-8879 x201

  2. #12
    Join Date
    Jun 2021


    Hello sky-knight,

    Thanks for the insightful responses.

    I solved it by remapping the interfaces so that WAN and LAN are only on the PCIe card ports 1 and 2. The MB NIC is unused for now.

    After that the speed was good (900/950) with 70% CPU usage (java).

    I don't really understand why it works and I guess you're right about saturation, but... happy ending

Page 2 of 2 FirstFirst 12

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

SEO by vBSEO 3.6.0 PL2