GL-MT300N-V2 - worth using as a router with SQM for 10 users

Technical Support for Routers
1

I bought this small router because I wanted to have some hands-on experience with openwrt and its SQM and the price was very good
Well, it seems to work fine but I did not do any high load throughput tests. As this product is generally advertised to be a travel router, I am not sure it is sufficient to serve approx 10 users simultaneously connected via ethernet cable (other wifi ap HW connected to the LAN port of MT300N). The users would mostly surf web and watch youtube videos and the WAN speed is 12/2megabit.

The reason I deployed this router there was I wanted to check the SQM which would reduce the bufferbloat problem. By doing short tests, it seems it works much better than without the SQM and surfing web is more responsive.

Should I consider buying more powerful hardware or will this small router have sufficient capacity for above solution?

2

SQM will almost always reduce available bandwidth. It can’t change the upstream buffering, only change it to “quick” local buffering to avoid the “slow” upstream buffering.

12 Mbps should be able to be managed by a relatively slow MIPS-based SoC.

I think the root of your problem is that 10 users on a 12 Mbps line is inadequate. 480p content is around 1.6 Mbps, 1080p content around 3.5 Mbps, so I wouldn’t expect to even get 6 users smoothly streaming 480p content with that line (or more than 2 users with 1080p content).

3

Hello,

I just made a quick and very “unscientific” test on my 300/300 Mbit fiber - and, You shouldn’t have any
problems with the CPU maxing out with 12Mbit Download and 2Mbit upload.

Using SQM, the CPU on my mt300-v2 maxes out and starts dropping a lot of packets at around 43-45 Mbit.Down/Up - this is without VPN or any other features running!

Remember to set SQM to around 80-85% of Your down/up bandwith (Bandwith = the WORST result You get from speedtest - this is important !)

Also remember to set the right “interface name” in Smart Queue Management (Basic Settings)- set it to “eth0.2 (wan, wan6)” - Default is ETH0.

When SQM is set to 10200 (download speed) 1700 (upload speed) and running a speedtest - the cpu load on the MT300-v2 fluctuates between 20 and 30 % (observed using htop)

So assuming you are only using the mt300-v2 for basic routing - You should absolutely be able to use it - NOT using QOS with 10 users on a 12/2 connection is disaster waiting to happen :slight_smile:

The only issue I can think of is the termal design of the unit - don’t know how it will perform when under constant load 24/7/365.

Br.

4

Thanks much for the replies and the test! I know the 12/2 is not suitable for 10 users but those my friends live with it for long time even without SQM (sometimes they complained about slow performance).
I wanted to give it a try using the SQM but I did not want to spend much money for expensive router as I was not sure whether it will actually help.

BTW what GL-inet router would be a better option with more power capacity? Or does GL-inet focus more into the travel routers for? I like openwrt and I do not want to experiment with brands which do not ship openwrt preinstalled. I saw for instance tp-links have several versions/hw revisions of the same model and it is not easy to know whether the one I will buy is fully supported by openwrt.

5

The B1300 and S1300 are the most powerful GL routers and are targeted for home use, so you should get one of those instead. All the other routers are like you say targeted for on the go use.

6

In my opinion, the router is not your limitation at this time, unless you have problems connecting over wireless. It is the 12 Mbps link.

The ipq40xx-based devices mentioned above would be very good choices at above 100 Mbps or so.

The wireless of the ipq40xx-based devices generally outperforms that of the inexpensive Mediatek (“MTK”, “GL-MT…”) devices or earlier-generation Qualcomm/Atheros (“QCA”, “GL-AR…”) devices.

7

That’s seems ridiculously conservative. The docs suggest 80-95% of a single measurement. The low end of that range already allows for pretty bad worst-case speeds.

I’d suggest doing a few measurements. Set SQM to 80% of the average then load the connection with a big upload and check to see how responsive other activities feel.

Next, set it to 95% and see how that is in comparison. If it “feels” the same, leave it. Otherwise, reduce it a bit and test again. Later, if things get permanently worse, complain to your ISP. If things get worse from time to time, then gradually drop the configured speeds until the bad times are rare enough to be tolerable for your typical use.

There is a tradeoff between max speed and variations in latency, but the relationship isn’t linear. You might be able to get from 20 VoIP glitches a minute to 2 by giving up a few % of maximum speed. Getting it down to 1 glitch a minute might require giving up another 10% (these numbers are just for illustrative purposes).

8

Just a couple of thoughts…

When using SQM - there is a fair amount of resource usage for directly connected clients, so more horsepower the better. AR300M has more to offer than the MT300N-V2, and it’s on a better understood chipset from an OpenWRT perspective.

Something worth looking at - using a device as a L2 “bump in the wire” approach. Very thoughful read below.

He’s using a D-Link DIR-825, and the AR300M has more compute and RAM resources available - and his ISP link is sub-100Mbit for his example.

One could do the same with something like a B1300 is links are above 100Mbit.

9

The IPQ40xx platforms (S1300/B1300) are a different class of performance compared to the Atheros/Mediatek MIPS devices.

That being said - I prefer the Atheros over MediaTek on the MIPS devices, mostly due to the wireless ATH9K drivers, as they’re well understood and relatively easy to work with for all WiFi STA modes.

10

I asked my friends how is the internet now (after I put there mt300n with sqm) and I got answer sometimes it seems a bit more responsive but sometimes it is as bad as it used to be. I was there and tried it just shortly couple of times - in my opinion it was better but sometimes really not ok - even a web page load took ages.
But now I am not sure what is the problem:

  • weak router not capable to process whole traffic with sqm
  • slow internet connection (12mb/2mb)
  • fluctuating speed of the internet connection (although it should be stable according to the isp)

The WAN connection is a local WIFI ISP who told me they checked the “line” and the signal quality is great and they can see the 12Mb download is almost fully utilized during the day. So I assume kids are on wifi and watch some youtube, whatever…

What was interesting - when I measured DL/UL speeds via dslreports, I was not able to get constant 12mb download. There were spikes with 12mb but no constant 12mb but I got average of ~10mb. And I measured this by having cable from the ISP wifi antena directly connected to my laptop. What’s worse, after I put GL-MT300N there (no other clients connected, just my laptop connected with ethernet cable to mt300n) the speed was slightly lower (maybe it was just a coincidence).

If it is mine internet connection at home, I would somehow diagnosed it, but I am in their place only from time to time so I can’t test it much. It can be also the ISP’s wifi connection is not so perfect as they claim and there are short intervals when the speed is very low (wifi is wifi… I do not know how many clients has their access point).

11

With SQM properly set, you will always have a loss of peak speed. SQM primarily deals with reducing delay (“buffer bloat” at the carrier), not increasing speed. It makes the line “smoother” by not pushing it to its limit.

The problem is almost certainly that there isn’t sufficient bandwidth to support 10 users on a 12 Mbps line with the speeds that they desire. There isn’t a router in the world that can magically increase bandwidth.

12

Hello :slight_smile:

Hopefully You understand what i write - english is not my first language :slight_smile:

@pepper - You wrote that you got an AVERAGE of 10mbit (download) using dsl reports - IF SQM is setup to assume a 12/2 Mbit connection (SQM set to 10200/1700 - 85% of 12/2 Mbit) however in real life, it is only getting 10/2 Mbit, then it vill not work properly once the connection is saturated.

If You use the AVERAGE result from speedtests taken during the day, then you will probably not notice any problems MOST of the time.

However if/when the bandwith drops “alot”, typically in the evening (if your ISP is overselling the bandwith) - then everything starts to suck, wife starts to get angry and children starts crying…

That is why i recommend using the WORST result from speedtests when setting up QOS - especially in “low-bandwith” scenarios. Remember to TURN OFF QOS when meassuring the bandwith, and do it at diffrent times throughout the day - also pay attention to the upload speed.

So, for fun i made some tests throtteling the bandwith to 10/2 KEEPING the (WRONG) SQM settings of 10200/1700 (85% of 12/2 mbit)

then pinged 1.1.1.1 (cloudflare) on a second PC - and got a stable baseping of 9ms

I then started to saturate the connection - and pings began to fluctuate between 9ms to 52ms and got a lot of “requests timed out” indicating that QOS was NOT working as it should anymore. I tried to open a copule of web-pages - they took ages to load.

when i set SQM appropriately to 8500/1700 (85% of 10/2), pings fluctuated between 9ms and 13 ms - and no “request timed out”, and the webpages loaded…slowly …

You Could try to set SQM to the above values, and see if that helps.

BUT as jeffsf said 12/2 is not much for 10 concurent users - a single 1080p Youtube video could use upto 5+ Mbit/s on average (Yes I know this is an oversimplification on how Youtube downloads work ) - So You are running out of bandwith very fast

B.r

oh don’t know if You have seen this video - demo starts at around 3:00