fix Bufferbloat on Edgerouter X (simply)

[Ceyarrecks]

{wow, lost all the text I just typed thinking the above links allowed one to preview in HTML} Excuse me Please: I need assistance in a simple means to configure the ERx v1.9.0 to resolve the below listed horrible test results. I have attempted to search these forums for "bufferbloat edgerouter x" but have failed to hone my understanding for what was found. I am new to the ERx environment, and my previous training and experience in similar technologies is failing me in my understanding. Thank you kindly for the assistance. CAH

[Ceyarrecks]

oh. forgot to re-type: Comcast is supposed to be providing a 15MB (or is it 25MB i forget now) connection.

[UBNT-Jordan]

You may want to try enableing smart queue on the QoS tab in the GUI and see if that helps with the buffer bloat. (also mentioned here)

[4thPotato]

This works fine for me:

In your case, you could start by setting upload rate to 1.5 Mbps and increase from there (1.6, 1.7 and so on) until you start seeing too much bufferbloat. This is assuming your upload is reasonably stable.

[idxman01]

Agreed, smart queue is working well for me.

What are your speeds if you bypass the router and go directly to your pc?

[Ceyarrecks]

Thank you kindly for the responses.

I will never a PC without the innate protection of a router, but I can say prior to having Comcast reducing the D/L speed, I recall having ~30MB down.

I will attempt to discern the picture for configuration(s)

Will post any further questions and/or results.

[Ceyarrecks]

I matched my ERx to your screenshot, set rate(s) to both 15, and 1.5, both failed with F's on bufferbloat.

http://www.dslreports.com/speedtest/6166112

[idxman01]

I know what you mean, it's a scary world out there.  I only do it for a minute during the test and it's something comcast will request before sending a tech.

[4thPotato]

Your first test result shows 4.81 / 2.17 Mbps. Your second test result shows 12.39 / 1.04 Mbps. As I said in my first post: "This is assuming your upload is reasonably stable.".

Obviously, your connection is not stable when doing these tests. Are you using wlan instead of wired when doing these tests? Do you have other people using the connection when doing these tests? If not, it's worth trying (as idxman01 suggested) to re-do the test without the er-x.

[Ceyarrecks]

the PC that runs the browser test is hard wired into a [switch] then [ERx] then [Motorola CM] booting a PC using a Live CD would probably be the only way I would connect a PC directly to the Cable Modem. so, a test I can do later today. while at this immediate moment, there is a laptop playing Mops radio and two other PCs running online games, these should not affect the test results, as I receive the same high bufferbloat when nothing is being run. so please forgive my inability to understand this topic, is bufferbloat due to a misconfiguration of a router, or something amiss on the ISP? I thought my understanding lent to a router problem, which is one of the reasons why I switched to the ERx from a Netgear running DDWRT.

[MertsA]

So I'd bet money that the buffering is happening on your cable modem. Typically cable modems are absolutely garbage at buffering because they have absolutely massive buffers that add a ton of latency when they fill up. The way to combat this, since you have no control over the cable modem, is to create a bottleneck before the device with massive buffers so that the bufferring happens on a device that you control and that has a reasonably sized buffer. What you typically do is run a speed test directly connected to the modem (you aren't going to get hacked by plugging directly into it, just get over it and do it) and then back off of that speed a little on the router. By setting the bandwidth on the router what you're doing is holding back traffic at the router so that it can't exceed whatever speed you specify. This is how we're making the router the bottleneck instead of the modem. The reason why this isn't working for you is because the router isn't the bottleneck with the speeds you specified, it's still the modem and it looks like you have a problem with your internet connection. If you want to test this you can set your download to 2Mbps and your upload to 1.5Mbps and run the speed test again and that quality will go up.

One thing to keep in mind is that if you're using a smart queue, just because the buffer is large on the router doesn't mean it presents the same problems as before. In the smart queue it'll split your traffic up into 1024 different queues. This is a good thing because it'll allow larger queues for high throughput traffic like filesharing where large queues smooth out bursts and allow for slightly closer to the maximum theoretical throughput but VoIP or web traffic will be put into a separate empty queue by itself most likely. The way it works is that the router picks one of the queues that has traffic in it and forwards the first packet. Normally with one big queue it's like having a huge line and all traffic goes to the very back of this line, having multiple queues allows traffic to skip the lines even if there's no prioritization going on and it just picks a queue using a round robin scheme because now each flow can only delay its own data. If you want, you can tune the smart queue so that each queue can only be a certain length to shorten the bufferbloat which is useful when dealing with low bandwidth connections as the default is going to be oversized for your connection. This won't cause too many problems as high throughput flows aren't really affected by latency for the most part and your low latency stuff is going to be in a separate queue that is unaffected but you can tune it to fix this if you want.

Unfortunately most of this is useless to you for the time being as you need to setup QoS such that the router is the bottleneck and right now, you can't do that without setting your bandwidth on the router to something that's below the minimum of what you get over a decent period of time. You would need to limit yourself to something like 3Mbps down or even less because of the problems with your internet connection which is pretty unusable. QoS works great when you have a stable connection and you get the same results from a speed test again and again. Go to 192.168.100.1 and you can see some of the details of your connection, in particular, go to the signal and connect tabs and let us know what's listed there. You mentioned it was a Motorola cable modem and it looks like you have Xfinity, what model cable modem are you using and do you own it or rent it?

[BranoB]

---

Ceyarrecks wrote:

I matched my ERx to your screenshot, set rate(s) to both 15, and 1.5, both failed with F's on bufferbloat.

http://www.dslreports.com/speedtest/6166112

---

What settings did you put in? Did you apply it to correct WAN interface?

In order for these settings to work, you have to first find out your real DL/UL speeds and apply the settings to match them.

[skempf]

I am having a similar problem. With no QoS, I have barely acceptable quality and huge bufferbloat. Using smart-queue, I have acceptable befferbloat and huge quality issues.

- no QoS http://www.dslreports.com/speedtest/6176090

- smart queue with defaults http://www.dslreports.com/speedtest/6176669

I am using an erl3, have an sb6121 cable modem, comcast economy 3/0.7 provisional speed connection.

@MertsAyou mention customizing the queue length for low bandwitdth connections. What setting would that be?

[botkiller]

I have 60Mbit down / 4Mbit up with provisioning typically at +10% and these are my settings and results. I read through many posts here and elsewhere to find what I understood to be more appropriate settings than the default for my connection. I only apply qos to the upload so the overall bufferbloat score is low. The real latency on the upload side is nearer to ~20ms than the 5ms default target. Setting target to 20ms and a few other tweeks made a big difference.

[Ceyarrecks]

here is m y results <20m old, from a laptop running Ubuntu Live CD directly to the cable modem. The cable modem was cold-booted. and the following was just received, with the ERx connected after cold-boot: one detail i do notice is the number of streams, 16 with laptop, 24 with router. I had previously MATCHED the first screenshot given, only changing the Rate as suggested, tried 1.5, and 15, neither changed the results received in those tests. (will continue to read the remaining posts to catch up)

[Ceyarrecks]

Thank you all, again, for the concerted effort assisting me in not only resolving, but understanding this issue I have.

what follows is from the Motorola Surfboard SB6141 Cable Modem Signal page:

{sorry for the huge size}

This page provides information about the current upstream and downstream signal status of your Cable Modem.

---

Downstream Bonding Channel Value

Channel ID	1	2	3	4	5	6	7	8
Frequency	651000000 Hz	657000000 Hz	663000000 Hz	669000000 Hz	675000000 Hz	681000000 Hz	687000000 Hz	693000000 Hz
Signal to Noise Ratio	39 dB	39 dB	39 dB	39 dB	39 dB	39 dB	38 dB	38 dB
Downstream Modulation	QAM256	QAM256	QAM256	QAM256	QAM256	QAM256	QAM256	QAM256
Power Level The Downstream Power Level reading is a snapshot taken at the time this page was requested. Please Reload/Refresh this Page for a new reading	12 dBmV	11 dBmV	11 dBmV	11 dBmV	11 dBmV	11 dBmV	10 dBmV	10 dBmV

Upstream Bonding Channel Value

Channel ID	18	17	19	20
Frequency	23700000 Hz	17300000 Hz	30100000 Hz	36500000 Hz
Ranging Service ID	14411	14411	14411	14411
Symbol Rate	5.120 Msym/sec	5.120 Msym/sec	5.120 Msym/sec	5.120 Msym/sec
Power Level	35 dBmV	35 dBmV	35 dBmV	35 dBmV
Upstream Modulation	[3] QPSK [3] 64QAM	[3] QPSK [3] 64QAM	[3] QPSK [3] 64QAM	[3] QPSK [3] 64QAM
Ranging Status	Success	Success	Success	Success

Signal Stats (Codewords)Bonding Channel Value

Channel ID	1	2	3	4	5	6	7	8
Total Unerrored Codewords	50234566	49074647	49092166	49083281	49067530	49068528	49069444	49070432
Total Correctable Codewords	13	94	1	0	17	41	14	13
Total Uncorrectable Codewords	1439	1503	1342	1373	1462	1422	1591	1569

---

[Ceyarrecks]

also, think I should add a bit of my history, which (Thankfully) includes where I had the higher cost Comcast connection which occurred in Nov2015; do these results show any insight?

History -  Local Time Down Up Ping B Q S = Link ISP Streams Button Tags Comment

2016-11-16 17:34:26	7.59	2.22	30	F	D		E	Cable	Xfinity	24 / 2
2016-11-16 17:23:10	11.93	2.33	16	F	F		F	Cable	Xfinity	16 / 3
2016-11-07 14:55:03	12.06	2.29	29	F	C		D	Cable	Xfinity	24 / 3
2016-11-07 14:50:26	1.44	2.18	29	F	C		D	Cable	Xfinity	24 / 3
2015-11-13 14:01:47	29.9	6.16	40	F	A+	D	C	Cable	Xfinity	24 / 6
2015-07-09 20:07:19	1.91	0.456	65	F		F		Wireless	Clearwire	12 / 1
2015-07-09 20:05:45	1.86	0.48	72	F		F		Wireless	Clearwire	12 / 1

[MimCom]

Those correctable/uncorrectable ratios do not look good.  Bad cable?  Bad splitter? Bad DOCSIS modem?

[Ceyarrecks]

this is an Internet only account.

there are no splitters.

there is one cable coupler to extend the cable drop ~10'

one detail I recall with installation, the tech stated is a "long run" as my apartment is exact opposite site of building from where pole tap is.

there is a new cable from pole to modem, as stated by the tech.

edit:

also, brand new cable modem. set behind UPS for electrical protection (no filtering on RG-6 cable)

editedit:

verified with Comcast the account is set to be 10MB/s download, 2MB/2 up.

[MimCom]

There is a splitter on the hardline that your drop connects to.  A moderately competent cable tech should be able to resolve this fairly quickly (assuming those error rates are outside their company standards.)  What do they say about the connection when you call in?  They can read all this stuff and test from their side.