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

Welcome To SNBForums

SNBForums is a community for anyone who wants to learn about or discuss the latest in wireless routers, network storage and the ins and outs of building and maintaining a small network.

If you'd like to post a question, simply register and have at it!

While you're at it, please check out SmallNetBuilder for product reviews and our famous Router Charts, Ranker and plenty more!

R7800 speed test shows high bufferbloat

Discussion in 'NETGEAR AC Wireless' started by Ernesto, Dec 6, 2018 at 6:23 PM.

  1. Ernesto

    Ernesto New Around Here

    Joined:
    Mar 26, 2017
    Messages:
    8
    Location:
    USA
    Hi,

    First post and a newbie.

    I just purchased a R7800 and put Voxel's latest firmware on the R7800. With the exception of bufferbloat everything seems to run like a champ. I am using DSLReports speed test and I get the speed that I have purchased from my ISP (100/10) but the bufferbloat seems high (sometimes reaching up towards 900 ms).

    I have only configured the basic stuff in the firmware: network ssid, network security. Is there something else I should adjust to reduce the bufferbloat.

    Thx,
    Ernesto
     

    Attached Files:

  2. EasternlightNZ

    EasternlightNZ New Around Here

    Joined:
    Thursday
    Messages:
    2
    Try turning off 11k on the router debug page?.

    My test http://www.dslreports.com/speedtest/42817602
     

    Attached Files:

  3. Ernesto

    Ernesto New Around Here

    Joined:
    Mar 26, 2017
    Messages:
    8
    Location:
    USA
    Thanks for the reply. I did that per
    e38BimmerFN
    comment on the Voxel's V1.0.2.61SF thread.

    The jpg I posted was with the 11K on the R7800 off.

    Also, the bufferbloat on uploads is very small (less than 30 ms). It's just the download speed that has the high bufferbloat values.
     
    Last edited: Dec 6, 2018 at 7:34 PM
  4. Ernesto

    Ernesto New Around Here

    Joined:
    Mar 26, 2017
    Messages:
    8
    Location:
    USA
    I ran some more tests with the DSLReports speedtest where I backed off the purchased download speed from my ISP. When I set the download speed to 75 (I purchased 100) I now get an A test result for bufferbloat from the speed test.

    Could this mean the router is doing what it's supposed to and as I approach the limit of my purchased bandwidth then bufferbloat increases? Could this be due to ISP limiting my download speed? Not sure there is anything I can do about that if it's a limitation on the download speed causing the bufferbloat to fail.
     

    Attached Files:

  5. XheliPilot

    XheliPilot New Around Here

    Joined:
    Oct 5, 2018
    Messages:
    9
    This artical helped me understand BB:


    Latency is more important to the speed of web page loads than link speed


    Most people think that their Internet line speed should be as high as possible so things load faster and work better. But that’s like thinking that if I have a car that can do 200MPH, I can cross Atlanta metro area three times as fast as a car that only does 65 MPH. The realities of traffic, stoplights, local congestion and many other factors will actually normalize the differences away. And one can even see a case where the slower top-speed vehicle could beat the other if it took a path with fewer lights and delays.
    So while top-speed is relevant in some scenarios, in the day to day, having fewer delays is much more beneficial.

    It’s the same on the internet, where having low latencies (less delay) can actually lead to a scenario where a 6Mbps DSL line can actually beat a 15Mbps line at loading a web page. Seriously, it can and does. Let’s see how.

    Remember that web pages create anywhere from dozens to hundreds of individual requests for information that make up a page. If each one of those has to wait a long time for a response, it doesn’t matter how ‘fast’ that response is downloaded at; the total time to get the page will be much longer. Let’s see an example using a 6Mbps line and a 15Mbps line:

    High Latency 15mbps line

    [​IMG]
    High Latency 15mbps line

    If the 15Mbps line has an average of 250ms of latency and a 5% packet loss rate, then the time to load the evenroute.com home page is 42 seconds.

    Low Latency 6mbps line

    " [​IMG]
    Low Latency 6mbps line

    Compared to a 6Mbps line with 23ms of latency and 3% packet loss, we can load the page in 7.9 seconds, or 5.3 times faster than the 15M line!

    And even if both lines had the exact same latency values, the ‘more than twice as fast’ 15Mbps line would load the page only 23% faster. So a true measure of ‘fast’ accounts for a mix of capacity and latency. More capacity alone will not make things go faster if you already have high latency.

    And some routers allow horrendous amounts of latency (also known as Bufferbloat), the example above used a conservative value, in real-life, we’ve observed averages of over 2,000ms (that’s 2 seconds worth), The maxes were in the 6,000 to 8,000 range. That’s more than six seconds lag on a given request!

    So the first and most urgent thing to address on any internet connection is the latency, as that has the biggest impact.

    As these diagrams illustrate, as latencies increase, page load time increases at a linear rate. However, increasing the download speed results in a non-linear rate of improvement in page load times. So controlling latency is more important than download speed in terms of both measured time and user perception of page loads.



    [​IMG]
    [​IMG]
    The good news is that with the right router, you can dramatically reduce those latencies down to manageable or even excellent levels. This results in pages that load quickly, smooth streaming and great Skype call quality.

    So ‘speed’ of internet interactions are heavily influenced by latency, and the IQrouter is the only router that works round the clock to analyze your line and adapt to it to deliver the lowest possible latencies and therefore highest quality and the quickest possible results at that




    Note that we are talking about the latencies while the line is loaded (which happens when big pages load or there is other traffic already on the link), many tests report the baseline ‘ping’ latency, and usually this is the ‘idle’ latency of your service. The lower this is the better, but not much under your control can impact that value. However, the latencies under load generated by the bufferbloat are controllable with the right router. So the numbers referenced above are based on average latencies from a bloated line and ISP
     
  6. Ernesto

    Ernesto New Around Here

    Joined:
    Mar 26, 2017
    Messages:
    8
    Location:
    USA
    I ran the speedtest without the R7800 in the mix (direct from cable moden to PC) and the bufferbloat results we unchanged. If I throttle the downlaod speed in the speedtest to 85% of the purchased bandwidth then bufferbloat is pretty low. I get the same results with the R7800 in the speedtest. So, I am not sure the R7800 is causing the bufferbloat in my case.

    I run a desktop connected to the router (or cable modem) via ethernet cable. The desktop operating system is Win 10 Pro 64 bit. Is there something else I should look at to minimize bufferbloat besides the R7800? Could there be an issue with the way Win 10 is configured? Or perhaps the cable modem is the bottleneck.

    I was going to try Lede openware or dd-wrt firmware but I don't think this will help since the bufferbloat results are the same with and without the R7800 in the testing.
     
    Flying Dutchman likes this.
  7. XheliPilot

    XheliPilot New Around Here

    Joined:
    Oct 5, 2018
    Messages:
    9
    That sounds about right. I am no expert but a docsis 3.0 modem is NOT built to control UL or DL speeds. This is done with the qos set up on your router. This is why when you lower your DL speeds on your router your BB score improves. Unfortunately you can't have your cake and eat it too...

    Plugging directly to your modem and testing should give you an F grade because no qos is being used to control the DL & UL speeds hence failing BB scores.

    The only modems that i am aware of that are using AQM which is supposed to help BB are the docsis 3.1 modems. Whether your isp supports these D3.1 modems is another question.. check out the netgear CM1000 it is backwards compatible with Docsis 3.0.
     
  8. Ernesto

    Ernesto New Around Here

    Joined:
    Mar 26, 2017
    Messages:
    8
    Location:
    USA

    Well, connecting my cable modem directly to the desktop yielded the same results as when the R7800 was between the modem and the desktop. Both setups failed on BB scores when the test was run at the purchased bandwidth. Both setups received A scores when the DL speed was throttled to 85% of the purchased bandwidth. Cable modem is an Arris TM822.
     
  9. XheliPilot

    XheliPilot New Around Here

    Joined:
    Oct 5, 2018
    Messages:
    9
    You do understand that you cannot get 100% of your "purchased bandwith" and an A for bufferbloat. You might get it close with some tweaking. Its one or the other. This would be the Holy Grail if we could get 100% of DL speed AND an A for BB.

    JFYI i pay for 300DL speeds and with QOS on i get 150DL MAX but i get A's & A+ across the board on dslreports.com.

    When i turn OFF QOS i get 310-320 DL speeds but... my Bufferbloat is an F.. as expected..
     
  10. Ernesto

    Ernesto New Around Here

    Joined:
    Mar 26, 2017
    Messages:
    8
    Location:
    USA
    I think we are saying the same thing. I get A scores on BB if I limit the test to 85% off purchase BW. So, I see no reason to jump to other firmwares in search of improvement (or the Holy Grial as you put it).
     
  11. XheliPilot

    XheliPilot New Around Here

    Joined:
    Oct 5, 2018
    Messages:
    9
    I have had ZERO issues with Voxels FW & it works for me. You are ahead of the game if you are getting A's for BB at 85% of your DL speeds.

    Im at half of my DL speeds to get an A for BB. I guess i have to do MORE tweaking lol.
     
  12. Rich Brown

    Rich Brown Occasional Visitor

    Joined:
    Apr 30, 2015
    Messages:
    14
    You're correct: DOCSIS 3.1 modems have an AQM (sometimes called "PIE") to manage bufferbloat. (DOCSIS 3.0 modems don't.)

    IQRouter from Evenroute is a commercial router that has AQM. (The "Latency is more important than the speed ..." article posted above comes from http://evenroute.com/latency-vs-speed/) The Ubiquiti routers also support AQM.

    You can also install open source firmware on your routers. OpenWrt firmware was the first to support AQM ("fq_codel" and now "cake"), and also supports the latest fixes for Wi-Fi latency and Airtime Fairness. These combine to decrease wireless latency (by orders of magnitude) and increase aggregate speed.
     
  13. Ernesto

    Ernesto New Around Here

    Joined:
    Mar 26, 2017
    Messages:
    8
    Location:
    USA
    I have a DOCSIS 3.0 telemodem.

    The results from doing speedtest directly from the modem to the desktop (R7800 out of the loop) were just as bad when I had the R7800 in the loop. So, if the telemodem has the same BB issues then how would installing LEDE on the R7800 improve the BB issue?
     
  14. Rich Brown

    Rich Brown Occasional Visitor

    Joined:
    Apr 30, 2015
    Messages:
    14
    The OpenWrt firmware (formerly LEDE) in the R7800 takes control of all traffic queueing, and simply doesn't allow the traffic queues to build up (even in your cable modem) so your latencies stay low. See the OpenWrt wiki on SQM (their name for AQM) for the details of setting it up.