1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

QoS for Tomato 1.23

Discussion in 'Tomato Firmware' started by venom, Apr 5, 2009.

  1. venom

    venom Addicted to LI Member


    I just switched from DDWRT v23 SP2 due to strange issues after a year of working great to Tomato v1.23.

    Everything on the Tomato works, and finally (and unlike DDWRT) the QoS seems to actually work. However, I am not getting the results I want.

    Just enabling QoS results in my bandwidth getting cut in half based on the results I entered for max and min DSL settings.

    I did a test download of a DVD from a Newsgroup. Normally when this happens my download speed is maxed out every time. So I will get about 525-560kb sustained. This also holds true for Tomato without the QoS enabled. When I enable the QoS and giving it 30 seconds to enabled, my speeds for that download go to 90kb and then slowly rise to a sustained 120kb.

    I have not entered any settings into the QoS other than my min/max up and download speeds based on DSL reports. Now, maybe I am getting KBytes mixed up with KBits?

    When I go to DSL reports my up is 550KB up (After 5 tests with no one grabbing any thing from me, and 5000 KB down. I entered those settings into the Tomato QoS. I also used a Bit Calc and the settings seem correct (I can usually upload to folks at about 70KB to 80KB so 550kbits = 68.75 kbytes)

    What might I be doing wrong here?

  2. bripab007

    bripab007 Network Guru Member

    Sounds to me like you either got the units wrong or swapped the values in downstream/upstream.

    Tomato is looking for kilobits per second, a.k.a. kbps, a.k.a. kbit/s. This is not to be confused with kilobytes per second, a.k.a. KB/s.

    Just about all the commercial speedtest websites (like this one: http://speakeasy.net/speedtest) display results in kilobits per second/kbps, as it sounds a lot more impressive than kilobytes per second and is also what most ISP advertise with.

    So, you DSL Reports test likely shows 550kbps up, not 550KB up, as you said (and 5000kbps down, not 5000KB). Take those an average of the results you see from of few of these speedtests (with QOS disabled, of course!), and then multiply that value by ~0.85 - 0.90, and enter the resultant value(s) into Tomato's QOS.
  3. venom

    venom Addicted to LI Member


    Thanks for the reply. Looks like you are definitely correct on the KB rather than kbps. However, I did enter the info verbatim from the site and confirmed them in Tomato QoS.

    550 Outbound
    5000 Inbound.

    I noticed the Default Class I set to medium which is set to 75-85% of my bandwidth. It still doesn't explain the massive speed loss in the Newsgroup download at all as I have no filters. Unless Bulk Data is altering this?

  4. bripab007

    bripab007 Network Guru Member

    Well, I'm not sure why I didn't mention this before, but, since you don't really have any real control over downstream QOS at your endpoint, you should just enter "999999" into the Inbound/Downstream rate box.

    That's the way I've always run my QOS, and it's worked great for years. Only in very extreme cases, in which you have lots of users or incredibly heavy P2P usage on your LAN do you need to worry about trying to shape the inbound rate.
  5. venom

    venom Addicted to LI Member

    Thanks again for the response. I have two roomates and one just got his XBox on here and all that XBox Live stuff. I don't want him hogging the bandwidth for when I game. I got both their MAC addresses but still, even after entering their info and setting it to Low, my download system bites (or bytes) it and the downloads are messed up.

    Let me try your suggestion and see what happens.

  6. az2008

    az2008 Addicted to LI Member

    I have my inbound QoS "max bandwidth" set to about 80-90% of my actual inbound speed. It makes a huge difference with VoIP (compared to setting it to unlimited). It's easily verified by making a test call while saturating my download capacity with www.speedtest.net. (I do it once with a valid max-bandwidth, and again with unlimited.). The difference in voice quality is night and day.

    If someone doesn't run an application as sensitive as VoIP, and/or they aren't maxing out their download speeds (like a speedtest does) maybe it wouldn't be as remarkable.

    The downside to setting "max bandwidth" (for inbound or outbound) is if a person has cable internet with "speed boost." Then it's hard to find the accurate sustained (actual) speed. And, if the max is set to that speed, the available "boost" speed will never be used.

    TCP Vegas is supposed to help with that. Theoretically a person can set inbound and outbound to unlimited. But, from what I can gather, TCP Vegas only influences TCP traffic, not UDP. And, only outbound traffic. However, the interrelationship between competing TCP and UDP traffic might be influenced by TCP Vegas slowing the outbound TCP traffic, letting UDP get more access. For different applications that might be useful (like VoIP).

  7. az2008

    az2008 Addicted to LI Member

    If the NNTP traffic isn't occurring on port 80, 443 or 53, it's undoubtedly falling into the "lowest" classification. You can see this by watching the chart/details while refreshing a newsgroup with a lot of posts. You can see how the connection is being classified.

    If you just want to throttle your roommate(s), I think I would delete all the classifications Tomato comes with. Just create classifications for MAC addresses.

    If you know there is certain traffic you want to give greater (or lessor) priority too, create more rules to target that traffic. Remember that the rules apply in the order they appear. If you want to reduce a only roommate's P2P music sharing more than their web browsing (not everyone's), you'd need multiple classifications with MAC address and other qualifications like L7 filter for P2P.


Share This Page