RT-N56U, 19.07.0-rc2, SQM performance

[jeniod]

I recently upgraded from 18.06 to 19.07 and noticed a significant performance degradation of luci-app-sqm.

In 18.06 i almost always got an 'A' bufferbloat rating @ http://www.dslreports.com/speedtest.
Now I get between 'C' and 'F'.

The issue seems to exist in both rc1 and rc2.

I have followed the guide https://openwrt.org/docs/guide-user/network/traffic-shaping/sqm, and tried fiddling with the settings, but nothing improves the performance noticeably.

I've monitored the Luci GUI Realtime Graph. It stops updating during a speedtest, and the peak value increases from ~2 to ~3.
When run 'top -n 1' via ssh, 'sirq' increases from always < 20 % to > 90 % and top also stop printing each second.

Is anyone else experiencing this issue?

Is the RT-N56U hardware getting too old for SQM / should I get a new router?

Any response is appreciated.

[darksky]

On a R7800 (dual core @ 1.70 GHz), I did not observe a difference between 18.x and 19.x-rc2. For me, downstream is CPU-limited to about 150 Mbps using cake/higher with another setting (see the thread I linked at the end of this post.)

Are you certain the change is related to the version? Note that bufferbloat tests should be run a few times to get an average value. See this thread.

[tmomas]

@jeniod

Now I get between 'C' and 'F'.

Last week I got B, today I got F, without any change in my network.
Repeat your test on different days.

[fantom-x]

@jeniod

I've monitored the Luci GUI Realtime Graph. It stops updating during a speedtest, and the peak value increases from ~2 to ~3.
When run 'top -n 1' via ssh, 'sirq' increases from always < 20 % to > 90 % and top also stop printing each second.

These are not impact-less and also consume a lot of CPU. Try your tests without any monitoring tools and logout of Luci.

[jeniod]

@darksky thanks for the thread link. Ill read it!
I tried doing the test a couple of times: D,D,D,C,A,B,F,D,D.
I looked at my history. From 18th to 28th September I did 17 tests with A or B grades.

[jeniod]

True. The tests it did in this reply was without monitoring tools, but I hadn't logged out. I just did a single test where I was logged out and got an F.

[jeff]

@jeniod

Is the RT-N56U hardware getting too old for SQM / should I get a new router?

https://openwrt.org/toh/asus/rt-n56u shows it as a 500 MHz, single-core, MIPS-based router from 2011.

A current, MT7628N running at 580 MHz caps out around 75 Mbps with SQM, without wireless, without anything else running, and on the bench. I don't know your line speeds, so I don't know if, say, 25-50 Mbps is "plenty" for you (my guess as to the RT-N56U's capabilities), or a limitation.

[jeniod]

Interesting numbers! Where did you find them?
Based on my experience and those numbers it does look like my router is getting too old for newer openwrt versions.

I'm running 50/50 Mbps, both 2.4 and 5GHz, and only added the SQM package.

I've limited the up/down in SQM to 40/40, 45/45, and 47.5/47.5. The bufferbloat does not change much, but the the up/down reported by dslreports increase when I increase the limits.

I can't disable the wifi right now, but I think it might be interesting to disable wifi and luci gui and run some more tests.

[jeff]

It’s not so much “later versions” but more “increased demands” that people are putting on their routers. A newer router will also give you 802.11ac.

Comparative Throughput Testing Including NAT, SQM, WireGuard, and OpenVPN

Overview While there's been a lot of individuals testing various routers, the question still always seems to come up, "What can I expect from Router X?" I won't say that my testing is going to definitively answer those questions, but at least it should be a self-consistent data set. To be very clear, this is a benchmark. Like most benchmarks, the performance you get in your own environment may be different. If you have a craptastic ISP, a lossy ISP connection, a flakey cable, hosts pummeling y…

[jeniod]

That sounds right. I should really get a new router

That's a really nice test. Thanks for the link!

However I cannot figure out what has changed in my environment except newer openwrt/sqm versions. I'm just curious to what's going on.

[fuller]

@jeniod

I'm just curious to what's going on.

rule of software developement:
if a regression costs less performance than one could gain by upgrading to the latest hardware, the regression is not fixed.
if it costs more, it is maybe fixed "if there is time available".

[darksky]

Actually, I may need to change my answer... When I have SQM enabled on 19.07.0-rc2, I am seeing much longer load times for webpages both on wireless devices and wired devices. When I disable SQM entirely, the slowness goes away. I will start a new thread so as not to hijack.
EDIT: and I will need a quantitative way to measure it.

[darksky]

...and here it is: What is a method to measure webpage lag?