Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
Next revisionBoth sides next revision
docs:guide-user:network:traffic-shaping:sqm [2022/01/14 15:56] – [Configuring the SQM Bufferbloat Packages] language cleanup part 2 palebloodskydocs:guide-user:network:traffic-shaping:sqm [2024/05/10 14:52] – remove a dead link, and small language tweaks palebloodsky
Line 1: Line 1:
 ====== SQM (Smart Queue Management) ====== ====== SQM (Smart Queue Management) ======
  
-OpenWrt has a pre-built package for controlling [[wp>Bufferbloat]] the undesirable latency that arises when the router buffers too much data.  +OpenWrt has a package called SQM for mitigating [[wp>bufferbloat]]the undesirable latency that arises when your router buffers too much data.
-OpenWrt calls the package Smart Queue Management (SQM), although it's also called [[wp>Active_queue_management|active queue management]] (AQM).+
  
-Bufferbloat is most evident when the link is heavily loaded. It causes bad performance for voice and video conversations, causes online games to lag, and generally makes people say, "The Internet is slow today."+Install ''[[:packages:pkgdata:luci-app-sqm|luci-app-sqm]]'' (or ''[[:packages:pkgdata:sqm-scripts|sqm-scripts]]'' if you don't use LuCI) and read below.
  
-The "luci-app-sqm" package of OpenWrt solves the problem of Bufferbloat. After a three-minute installation and configuration, you'll have a much more lively network connection. Here's how:+===== Overview =====
  
-**TL;DR** Video guide: [[https://www.youtube.com/watch?v=FvYhifdQ92Q]]+Bufferbloat is most evident when a connection is heavily loaded with downloads or uploadsIt causes increased latency (ping), resulting in poor performance for realtime apps like VoIP, video chat, lag in online games, and generally makes the internet feel unresponsive. This can be mitigated with SQM.
  
-Advanced SQM tuning guide: [[https://www.bufferbloat.net/projects/bloat/wiki/Getting_SQM_Running_Right/]]+SQM is an integrated system that performs per-packet/per-flow network scheduling, active queue management (AQM), traffic shaping, rate limiting, and QoS prioritizationIn comparison, “classic” AQM only manages queue length and “classic” QoS only does prioritization.
  
-**Note**: SQM is heavily CPU-basedOlder or slower devices [[https://forum.openwrt.org/t/so-you-have-500mbps-1gbps-fiber-and-need-a-router-read-this-first/90305|may not be able to keep up with your connection]]. +SQM is heavily CPU dependentSlower devices may not be able to [[https://forum.openwrt.org/t/so-you-have-500mbps-1gbps-fiber-and-need-a-router-read-this-first/90305|keep up]] with your connection speed.
-===== Preparation: Measure Your Current Speed and Latency ===== +
  
-Before you can optimize your network, you need to know its current state. Run a speed test to find your down/upload link speeds. To do this: +SQM is incompatible with software/hardware [[https://www.redhat.com/en/blog/should-i-offload-my-networking-hardware-look-hardware-offloading|flow offloading]] which bypasses part of the network stackBe sure to uncheck those features in LuCI -> Firewall to use SQM.
-/* I removed the discussion of the very cool speedtest-netperf package because it's a distraction from getting SQM configuredWe should advise readers simply to use DSLReports... - richb-hanover */ +
-/+
-To install the //speedtest-netperf// package: +
-  * From LuCI web GUI:  +
-    * choose **System -> Software**, then filter on the name **speedtest-net** and click the **Install..** button. +
-  * Or, from cmd: <code>opkg update && opkg install speedtest-netperf</code>+
  
-To use the script, ssh to the OpenWRT device and run the following (example results shown, additional options are available): +===== Preparation: Measure Your Current Speed and Latency =====
-<code> +
-speedtest-netperf.sh -H netperf-west.bufferbloat.net -t 20 +
-2020-08-25 14:35:40 Starting speedtest for 20 seconds per transfer session. +
-Measure speed to netperf-west.bufferbloat.net (IPv4) while pinging gstatic.com. +
-Download and upload sessions are sequential, each with 5 simultaneous streams. +
-..................... +
- Download: 805.68 Mbps +
-  Latency: [in msec, 21 pings, 0.00% packet loss] +
-      Min:  19.125 +
-    10pct:  21.901 +
-   Median:  29.465 +
-      Avg:  31.586 +
-    90pct:  36.378 +
-      Max:  63.253 +
- CPU Load: [in % busy (avg +/- std dev) @ avg frequency, 18 samples] +
-     cpu0:  89.4 +/-  4.7  @ 1415 MHz +
-     cpu1:  31.3 +/-  2.7  @  848 MHz +
- Overhead: [in % used of total CPU available] +
-  netperf:  27.2 +
-..................... +
-   Upload:  600.61 Mbps +
-  Latency: [in msec, 21 pings, 0.00% packet loss] +
-      Min:  19.627 +
-    10pct:  19.683 +
-   Median:  21.432 +
-      Avg:  21.521 +
-    90pct:  22.424 +
-      Max:  26.307 +
- CPU Load: [in % busy (avg +/- std dev) @ avg frequency, 19 samples] +
-     cpu0:  89.0 +/-  5.5  @  1401 MHz +
-     cpu1:  33.5 +/-  4.0  @  890 MHz +
- Overhead: [in % used of total CPU available] +
-  netperf:   2.5 +
-</code> +
-*/ +
-  * When your network is relatively quiet, use DSLReports Speedtest at: http://www.dslreports.com/speedtest You will need this information in the next step. (See [[https://forum.openwrt.org/t/sqm-qos-recommended-settings-for-the-dslreports-speedtest-bufferbloat-testing/2803]] for how to get the most out of the DSLReports Speedtest) +
-  * A decent alternative to the dslreports test that also helps to asses bufferbloat is waveform's bufferbloat test [[https://www.waveform.com/tools/bufferbloat]]; while you are running that test, waiting for the results, their "Frequently Asked Questions" section is worth reading... the biggest relevant difference to the dslreports test is waveform uses traffic sources and sinks in cloudfare's CDN which should be closer to most end-users. +
-  * //This is probably a good time to [[docs:guide-user:troubleshooting:backup_restore|back up your configuration]]. (Optional, but always a good idea.)// +
-   +
-===== Configuring the SQM Bufferbloat Packages =====+
  
-The default values will work well and represent conservative estimateswhich is desirable compared to under-estimatingYou should be able to improve performance further by setting values below.+Before you can optimize your networkyou need to know its current stateWhen your internet is quiet, run a speed test to find your peak download/upload speeds: 
 +  * Run a speed test from [[https://www.waveform.com/tools/bufferbloat|Waveform]] or [[https://speedtest.net|Speedtest]]. Both display the latency during download and upload traffic, and grade your existing bufferbloat. 
 +  * If you are using this OpenWrt device as an [[docs:guide-user:network:wifi:relay_configuration|Extender, Repeater, or Bridge]], test your upstream router (OpenWrt or otherwise) and determine if an issue is present there first.
  
-To configure SQM, go to **Network -> SQM QoS**.+===== Configuration ===== 
 + 
 +In LuCI go to **Network -> SQM QoS**. The default settings will work, however you can improve performance with settings specific to your internet connection as described below:
  
   - In the **Basic Settings** tab:   - In the **Basic Settings** tab:
     * Check the **Enable** box     * Check the **Enable** box
-    * Set the **Interface name** to your wide area network (WAN) link. Interfaces are listed in the dropdown, or check **Network -> Interfaces** to find the name for the WAN port. +    * Set the **Interface name** to your internet (WAN) link. Interfaces are listed in the dropdown, or check **Network -> Interfaces** to find the WAN port. 
-    * Set the **Download** and **Upload** speeds to 85-95% of the speed you measured in Preparation +    * Set the **Download** and **Upload** speeds to 90% of what you measured in Preparation 
-  - In the **Queue Discipline** tab, you can leave the default settings or set the following+  - In the **Queue Discipline** tab: 
-    * Choose //cake// as the Queueing Discipline+    * Choose //cake// as the Queueing Discipline (or //fq_codel//, consider [[:docs:guide-user:network:traffic-shaping:sqm#a_little_more_sqm_tuning|note 2]])
     * Choose //piece_of_cake.qos// as the Queue Setup Script     * Choose //piece_of_cake.qos// as the Queue Setup Script
-    * Advanced Configuration may be left unchecked. +    * Advanced Configuration may be left unchecked 
-  - In the **Link Layer Adaptation** tab, choose the kind of link you have (optional, default may also be used): +  - In the **Link Layer Adaptation** tab, select your link type (optional: set mpu see [[:docs:guide-user:network:traffic-shaping:sqm#a_little_more_sqm_tuning|note 3]]): 
-    * //For VDSL// - Choose **Ethernet**, and set overhead 34 (or 26 if you know you are not using PPPoE) +    * //For VDSL// - Choose **Ethernet**, and set overhead 34 (or 26 if you're not using PPPoE) (mpu 68). If the link uses 100 Mbps ethernet, set overhead 42 (mpu 84). 
-    * //For DSL of any other type// - Choose **ATM**, and set overhead 44 +    * //For DSL of any other type// - Choose **ATM**, and set overhead 44 (mpu 96). 
-    * //For DOCSIS Cable// - Choose **Ethernet**, and set overhead 22 +    * //For DOCSIS Cable// - Choose **Ethernet**, and set overhead 22 (mpu 64). For rates > 760 Mbps, set overhead 42 (mpu 84), because 1 Gbps ethernet between modem and router affects the worst-case per-packet-overhead. 
-    * //For true ethernet or Fiber to the Premises// - Choose **Ethernet**, and set overhead 44 +    * //For Ethernet or Fiber// - Choose **Ethernet**, and set overhead 44 (mpu 84). 
-    * //When in Doubt, it's better to overestimate// - Choose overhead 44 +    * //If unsure, it's better to overestimate// - Choose overhead 44 (mpu 96). 
-  - Click **Save & Apply**. That's it!+  - Click **Save & Apply**. 
 + 
 +That's it! You can confirm mitigation of bufferbloat by re-running the speedtest. Any increased ping during download/uploads will now be minimal. 
 + 
 +===== Results ===== 
 + 
 +As an example, the user below is running OpenWrt 23.05 on a [[toh:linksys:wrt_ac_series|WRT32X]] router. The internet connection is a DOCSIS cable modem with 500/35 Mbit service. Note this ISP includes over-provisioning. Cake was selected with 90% dl/ul limits on baseline speedtest values. Latency increase under load dropped to zero, lower ping with no packet loss is observed during VoIP and online gaming during heavy internet usage. Speedtests results with and without SQM:
  
-Measure your latency again with the speed test. You should notice that the measured ping times are only slightly higher during the downloads and uploads. www.dslreports.com/speedtest should now report A+ ratings for Bufferbloat and Quality. Latency sensitive applications like VoIP, Skype, Facetime, gaming, or general web browsing will be much more consistent even if someone'uploading or downloading a lot of data.+[[https://www.waveform.com/tools/bufferbloat?test-id=e101a8fc-f017-4eef-8f90-b27bcb783d62|User'speedtest results with SQM.]]
  
-You've reduced your connection's bufferbloat!  +^   Speedtest Results   ^^^^^^^^ 
-===== A Little More SQM Tuning  =====+| QoS  | Download  | Upload   | Unloaded Ping | DL Latency | UL Latency | Quality grade | Bufferbloat grade | 
 +| None | 532 Mbits | 37 Mbits |  12 ms        | +18 ms     | +38 ms     | B             | B                 | 
 +SQM  | 495 Mbits | 28 Mbits |  12 ms        | +0 ms      | +0 ms      | A+            | A +               |
  
-The steps above will control latency well without additional effort. The 80-95% range mentioned above is a good starting point, but you can often gain more speed while still controlling latency by making a couple experiments to adjust the settings.+===== A Little More Tuning  =====
  
-If you want to spend a few more minutes tuningdo these steps.+1. The steps above will handle latency wellbut you may improve this further via adjusting the dl/ul limits and with these steps:
  
-  * Increase the Download speed until the latency begins to increase, then go back to a slightly lower value.  +  * Increase the Download speed limit and retest until latency begins to increase, then go back to a slightly lower value. 
-  * Do the same for the Upload speed entry.  +  * Repeat the same for the Upload speed. 
-  * It may be worth your time to tweak the two a bit up and down to find a sweet spot for your connection and usage. +  * Use [[https://www.waveform.com/tools/bufferbloat|Waveform]] speedtest to achieve A+ quality and A+ bufferbloat grades when optimal settings are found
-  * We recommend you use [[http://dslreports.com/speedtest|DSLReports Speed Test]] for the latency tests because it measures both speed and latency at the same time.  +  * For DSL, the experiments above may produce download/upload values that are actually //higher// than the original speed test. This is ok, the ATM framing bytes of a DSL link add an average of 9% overhead, and these settings tell SQM how to make up for that overhead
-   +  * For DOCSIS cable, some providers trick speed tests by adding 10-15% over-provisioning for the first 10 seconds (so speed tests look better!).
-**Note 1:** If you have a DSL link, the experiments above may produce Download and Upload values that are actually //higher// than the original speed test results. This is OK: the ATM framing bytes of a DSL link add an average of 9% overhead, and these settings simply tell SQM how to make up for that overhead.+
  
-**Note 2:** If you use a cable modemyou should use speed test that runs for a longer time. Cable modem makers have tricked speed tests by using "speedboost"which can provide an extra 10 Mbits or so for the first 10 seconds (so the speed test will look good(!))Don'be surprised if the "right" setting for your queue rates is lower than the no-SQM speed test results. You may need to tune the speeds //down// from your initial settings to get the latency to the point you need for your own usage of your connection.+2. Cake is the [[https://www.bufferbloat.net/projects/codel/wiki/Cake/|preferred algorithm]] as it is excellent at mitigating bufferbloat. HoweverFq_codel is often fasteralbeit less comprehensive optionOne user found [[https://forum.openwrt.org/t/netgear-r6220-sqm-results-downstream-cut-in-half-and-my-optimal-settings/114301|fq_codel gave about 15% higher throughput when CPU limited]] and this [[https://lists.bufferbloat.net/pipermail/cake/2018-April/003384.html|email thread showed similar results]].
  
-**Note 3:** You can also experiment with the other settings (read [[docs:guide-user:network:traffic-shaping:sqm-details|SQM - The Details]]  for more information), but they will not make nearly as large a difference as ensuring that the Download and Upload speeds are maximized+3. To set your link **mpu** (read [[docs:guide-user:network:traffic-shaping:sqm-details|SQM Details]] and [[https://forum.openwrt.org/t/sqm-setting-question-link-layer-adaptation/2514/9|SQM setting question]]for efficiency improvements. Setting mpu will ensure rate shaping is correct for small packets.
  
-**Note 4:** Check the [[docs:guide-user:network:traffic-shaping:sqm-details#faq|FAQ]] and [[docs:guide-user:network:traffic-shaping:sqm-details#troubleshooting_sqm|Troubleshooting SQM]] guides.+4Check the [[docs:guide-user:network:traffic-shaping:sqm-details#faq|FAQ]] and [[docs:guide-user:network:traffic-shaping:sqm-details#troubleshooting_sqm|Troubleshooting SQM]] guides for more information. See also: [[:docs:guide-user:network:traffic-shaping:sqm_configuration|SQM configuration]] for advanced options.
  
-**Note 5:** Why Cake is typically the preferred algorithm [[https://www.bufferbloat.net/projects/codel/wiki/Cake/]] as it is more CPU and RAM efficient than fq_codel. Or at least it was early in its development, currently the argument is more that cake bundles almost all features that are helpful for a typical private internet access linkOne user found [[https://forum.openwrt.org/t/netgear-r6220-sqm-results-downstream-cut-in-half-and-my-optimal-settings/114301|fq_codel gave about 10% more performance on a slow CPU.]]+5. Consider Cake tuning parameters: [[https://man7.org/linux/man-pages/man8/tc-cake.8.html]]
  
-**Note 6:** Read more about Cake tuning on the man page including link layer adaptation settings: [[https://man7.org/linux/man-pages/man8/tc-cake.8.html]]+6. To use SQM it is necessary to disable hardware flow offloading as per [[https://forum.openwrt.org/t/sqm-and-non-sqm-queue-issues-lan-vs-wlan/15433/10|this post]].
  • Last modified: 2024/11/07 18:49
  • by palebloodsky