Understanding QOS statistics

Discussion in 'Tomato Firmware' started by davexnet, May 6, 2008.

  1. davexnet

    davexnet LI Guru Member

    Hello, I'm using the Tomato firmware for the first time in my Wrt54gl.
    It's installed and working fine. I've attempted to configure the QOS -
    haven't changed much except up and down bandwidth on "basic settings".
    Download 5700, upload 492 Kbits.

    I have left the setting in "classification" as-is.

    When I look at "view graphs" bandwidth distribution, I see 20% is allocated to
    "highest" using approx 100 Kbits per second. But in the "contents distribution",
    "highest" has 0 connections.

    I'm using Utorrent and it's currently DL'ing at 200KBytes PS
    and uploading at 43 KBytes per second. Also, concurrently downloading from a
    newsserver at approximately 350 KBytes P/S. (I can deduce this from the
    realtime bandwidth display) Browsing the web has slowed down a little, but
    still responding.

    Any enlightenment appreciated.
  2. davexnet

    davexnet LI Guru Member

    Another thing,
    assuming you set your upload limit in the firmware properly,
    (say 90% of actual), what happens when the data coming into the router
    exceeds that amount ? For example, you specify 60 Kbytes
    as the limit in the firmware, but in the p2p application 70 Kbytes is set as
    the upload limit. (I know it's not a great example, but I just need something
    to illustrate my point)

    What does this do to QOS ? Presumably, the router is saturated, more than it can
    send to the modem - what's the effect of this ?

    On the other hand, if I set all my uploading apps to only upload 50Kbytes
    aggregate, it leaves 10 KBytes spare (based on the above setting of 60) -
    plenty for web browsing - then why do I need QOS at all ?

    Do you see my point ? if the applications saturate the router,
    is QOS still working ?

    Thanks for any info.
  3. unicorn02

    unicorn02 LI Guru Member

    I also observe this. Highest also on my tomato has the most percentage allocated, but shows zero connections. I would expect to see the traffic in category "A", because this was the only category active at my tests, but instead "a" shows only 7%. Is this a bug?

    I mean that highest shows a large percentage of the traffic, but actually no associated connections? I even did not have categorized any traffic as highest. So highest is not associated to any rule in my router....

    Update: I found out that the "highest" categorization is being used for Priotized ACKs if this option is selected in tomato! If I deselect this option, then I get the traffic "correctly" shown as expected.

    So the ACKs are the "invisible" highest priority traffic.
  4. rhester72

    rhester72 Network Guru Member

    If you are prioritizing ACKs or the like, those will show up in Highest even though the connection itself may be associated with another category.

  5. bripab007

    bripab007 Network Guru Member

    You should just enter "999999" in your inbound bandwidth (a.k.a. downstream/download bandwidth) box in QOS because Tomato performs no QOS on incoming packets, as it's not an effective way to shape traffic. Thus, you are just artificially limiting your inbound bandwidth.
  6. az2008

    az2008 Addicted to LI Member

    I can say from experience that setting my inbound "max bandwidth" correctly makes a *huge* improvement on my VOIP performance.

    The whole thing about whether inbound "QoS" should be called "QoS", or if delaying "acks" and dropping packets (in an attempt to slow down the sender) is a good idea. But, all I can say is that it seems very effective to me.

  7. bripab007

    bripab007 Network Guru Member

    Interesting...I don't use VOIP, but I regularly vid/voice chat and play on Xbox Live, and setting inbound to "unlimited" bandwidth has always worked well for me...minimum of dropped/delayed packets, according to the limited testing, as well as years of real-world usage, have shown me.
  8. az2008

    az2008 Addicted to LI Member

    My connection is relatively slow (1500kbs/250kbs). Could that be a difference.

    All I know is that it makes a big difference for me when I use VOIP and go to speedtest.net, and start a test which saturates my connection.

    I can't say I've tried 999999. I think that would be similar (for inbound) as disabling QoS. I'll have to try it so I know for sure.

  9. unicorn02

    unicorn02 LI Guru Member

    thanks rodney, just found that out by myself short time ago.
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice