Weird KH Latency?

Discussion in 'The Lounge' started by Caleb, Apr 1, 2007.

  1. Caleb

    Caleb McDonald's Drywall Expert

    Yeah I was just messing around and noticed something peculiar...


    When I ping knownhost.com I get 38-40ms
    When I ping vz01-tx, vz02-tx, etc. I get 48-51ms
    When I ping one of my sites, I get 47-55ms
    When I ping virpus.com (also in C4D) I get 42-45ms

    Here are some traceroutes...


    $ traceroute vz01-tx.privatesystems.net
    traceroute to vz01-tx.privatesystems.net (65.99.235.201), 64 hops max, 40 byte packets
    1 192.168.1.1 (192.168.1.1) 2.785 ms 2.475 ms 3.134 ms
    2 10.20.64.1 (10.20.64.1) 19.580 ms 14.656 ms 11.363 ms
    3 12-215-20-193.client.mchsi.com (12.215.20.193) 12.758 ms 12.435 ms 10.291 ms
    4 12-215-0-74.client.mchsi.com (12.215.0.74) 22.253 ms 15.866 ms 15.472 ms
    5 tbr2-p011604.sl9mo.ip.att.net (12.122.82.26) 26.020 ms 19.579 ms 22.572 ms
    6 ggr2-p310.cgcil.ip.att.net (12.123.6.65) 25.313 ms 17.220 ms 19.537 ms
    7 192.205.33.186 (192.205.33.186) 30.605 ms 17.399 ms 21.159 ms
    8 ae-1-51.bbr1.chicago1.level3.net (4.68.101.1) 16.652 ms ae-1-55.bbr1.chicago1.level3.net (4.68.101.129) 17.597 ms 17.644 ms
    9 as-0-0.bbr1.dallas1.level3.net (209.247.11.22) 48.485 ms 49.820 ms ae-0-0.bbr2.dallas1.level3.net (64.159.1.110) 122.097 ms
    10 ae-21-52.car1.dallas1.level3.net (4.68.122.45) 55.893 ms 48.564 ms 49.212 ms
    11 colo4-dalla.car1.dallas1.level3.net (8.9.232.74) 48.114 ms 49.415 ms 49.159 ms
    12 ge-1-2.core2.colo4dallas.net (206.123.64.30) 54.092 ms 50.932 ms 52.394 ms
    13 vz01-tx.privatesystems.net (65.99.235.201) 50.988 ms 50.117 ms 50.379 ms


    And


    $ traceroute knownhost.com
    traceroute to knownhost.com (65.99.213.9), 64 hops max, 40 byte packets
    1 192.168.1.1 (192.168.1.1) 5.567 ms 20.493 ms 2.641 ms
    2 10.20.64.1 (10.20.64.1) 10.100 ms 12.468 ms 9.461 ms
    3 12-215-20-193.client.mchsi.com (12.215.20.193) 27.922 ms 12.233 ms 11.591 ms
    4 12-215-0-74.client.mchsi.com (12.215.0.74) 17.997 ms 19.464 ms 16.357 ms
    5 tbr2-p010201.cgcil.ip.att.net (12.123.6.78) 45.042 ms 49.540 ms 44.344 ms
    6 tbr2-cl7.sl9mo.ip.att.net (12.122.10.46) 45.720 ms 49.794 ms 47.027 ms
    7 tbr1-cl24.sl9mo.ip.att.net (12.122.9.141) 45.375 ms 40.727 ms 50.362 ms
    8 tbr2-cl6.dlstx.ip.att.net (12.122.10.90) 43.077 ms 54.755 ms 44.386 ms
    9 12.122.100.69 (12.122.100.69) 43.915 ms 47.858 ms 43.253 ms
    10 * * *
    11 border4.tge4-1-bbnet2.ext1.dal.pnap.net (216.52.191.94) 59.908 ms 40.812 ms 41.395 ms
    12 colo4dallas-3.border4.ext1.dal.pnap.net (216.52.189.10) 45.598 ms 47.051 ms 52.418 ms
    13 206.123.64.38 (206.123.64.38) 47.829 ms 48.520 ms 46.363 ms
    14 vz01-tx.privatesystems.net (65.99.235.201) 44.021 ms 53.735 ms 44.282 ms
    15 knownhost.net (65.99.213.9) 57.137 ms 40.795 ms 40.842 ms


    I've always gotten about 38-40ms to my VPS at knownost, so what's up? I guess I'm a bit picky but 10-12ms is about 25% more than what it should be. Let me know if anybody else sees this.


    -Caleb
     
  2. KH-Paul

    KH-Paul CTO Staff Member

    Hi Caleb,

    Above mentioned traceroutes are going through different providers - first route went through Level3, second through InterNap. Also reverse route (in case of asymmetric routing) is important too. It is up to the transit provider which route to select for specific network / route object and this is not something that can be managed from our side.
    Other points that may affect ping time are:
    - icmp packet processing priority. All our equipment is set to process ICMP packets with lowest possible priority. There is no reason to spend time processing ICMP packets if real traffic is going through. ICMP packets are pushed to the end of the input / output queue on all our devices;
    - Node location. We run our servers on gigabit network with multiple Cisco switches connected to each other. RTT might depend on how many internal switches your packet needs to go through. Also, for such "picky people" physical distance between cabinets may also matter as it will add some fraction of millisecond while packet will travel from the cabinet where uplink from Colo4Dallas is coming to to the cabinet where your physical node is located ;)

    PS knownhost.com is not any different from any other VPS. It runs inside standard VPS account based on VPS-L plan without any additional tweaks.

    Regards,
    Paul
     
  3. Caleb

    Caleb McDonald's Drywall Expert

    Hey, one more thing. I have always been confused with latency.

    centi = hundred (as in, per CENT)
    milli = thousand (1000mL = 1L)

    so is 40ms 2/5 of a second, or 1/25 of a second? If m in ms stands for milli, then it should be 1/25, but I know that's not really possible.


    Thanks
     
  4. KH-Paul

    KH-Paul CTO Staff Member

    Caleb,

    ms = millisecond. 40ms = 40/1000 of a second or 0.04s. 2/5 of a second would be 400ms.

    Regards,
    Paul
     
  5. ppc

    ppc Moderator

    A standard VPS-L package is able to handle all the traffic and demand of Knownhost.com?!
     

Share This Page