My QoS settings are:
Port: WAN
Packet Scheduler: HTB
Queuing Discipline: CAKE
Downlink: 330000
Uplink: 11500
Rest of settings default nothing else changed
Firmware: DD-WRT v3.0-r49567 std (07/27/22)
I have 300/10 Cable, and just got bumped up from 100/10 cable. Before my Downlink cap was 110000. I've tried other down and up speeds and also queuing disciplines and this seems best. Am I right? I really haven't read anything that helps me here from googling it and looking on the forums. Am I at the end of the road and set up properly? My setup seems to not take too much bandwidth away and keeps me below 15ms on the up and down when maxing it out with buffer bloat tests. Or am I full of it? I'd be very interested in getting some input. Thanks. If it helps I don't bother using the wifi in the router it's off, I have TP-Link routers in AP mode for my AP's.
I don't currently use QoS, but from memory you have to actually choose speed caps for the worst expected case. For example, I used to have a provider with service that would significantly slow down at different times of day and days of week. Thus, I had to set the QoS caps to those slowest measurements rather than the best or advertised measurements, especially since it was at those times that QoS was actually needed! Basically, QoS settings are really just trial and error so no one here can tell you what's optimal unless thay happen to be your next door neighbor with identical service. _________________ My DD-WRT Routers:
Linksys WRT3200ACM - Marvell
Linksys WRT1900ACS - Marvell
Netgear R9000 - Atheros
Netgear R7000 - Broadcom
PC x86-64 VM - Atheros
I don't currently use QoS, but from memory you have to actually choose speed caps for the worst expected case. For example, I used to have a provider with service that would significantly slow down at different times of day and days of week. Thus, I had to set the QoS caps to those slowest measurements rather than the best or advertised measurements, especially since it was at those times that QoS was actually needed! Basically, QoS settings are really just trial and error so no one here can tell you what's optimal unless thay happen to be your next door neighbor with identical service.
Sounds like you think I'm "at the end of the road and setup properly" I take it? I use QoS because my cable internet connection gets a hefty latency spike once I start getting to download and upload values close to the max speed. Whenever the connection throttles up to deliver all my speed the latency is really bad unless I use QoS.
Does not sound to me like a correct setting unless your connection is extremely overprovisioned.
The QoS data rates are set to about 85-90% of the maximum measured throughput.
If you have set 330Mbit in the downlink, the connection should normally offer a download speed of 366-390Mbit without QoS.
If you measure only 300Mbit without QoS then you should set the QoS download rate to about 255-270Mbit.
Otherwise, without precise measurements, no one can tell you whether you have adjusted it well or badly.
You have only indicated that you have a 300/10 cable connection.
This doesn't help us because it doesn't reflect the real data rates and actually doesn't match your settings.
And I believe you that the connection has a high bufferbloat.
This is a bad cable connection with a high asymmetry in the ratio of 30:1.
Even the bad DSL connections have only a ratio of 15:1 (15Mbit down 1Mbit up).
Does not sound to me like a correct setting unless your connection is extremely overprovisioned.
The QoS data rates are set to about 85-90% of the maximum measured throughput.
If you have set 330Mbit in the downlink, the connection should normally offer a download speed of 366-390Mbit without QoS.
If you measure only 300Mbit without QoS then you should set the QoS download rate to about 255-270Mbit.
Otherwise, without precise measurements, no one can tell you whether you have adjusted it well or badly.
You have only indicated that you have a 300/10 cable connection.
This doesn't help us because it doesn't reflect the real data rates and actually doesn't match your settings.
And I believe you that the connection has a high bufferbloat.
This is a bad cable connection with a high asymmetry in the ratio of 30:1.
Even the bad DSL connections have only a ratio of 15:1 (15Mbit down 1Mbit up).
300/10 is what the ISP calls it. That's just how it is in my region, sir. My down and up are the highest values I can set with the latency not boosting up more than 15ms during bufferbloat testing, that's why they are set that way. Setting them higher gives me latency during bufferbloat test, setting them lower gives no latency improvement during bufferbloat test and lower bandwidth. The CAKE setting does seem to work fine for me. What I'm kind of digging at I guess is that there is no good dd-wrt focused faq or guide as far as explaining all this, I see some things saying CAKE doesn't work or work right on dd-wrt and it seems to me it does...I'm probably fine and don't need to change anything I'm starting to think.
Joined: 08 May 2018 Posts: 14126 Location: Texas, USA
Posted: Sun Jul 31, 2022 21:27 Post subject:
I have a 150/10 Cable internet connection. That much "difference" between downstream and upstream is "normal". I have a different modem/router and not sure if I can pull up the numbers, but it depends on number of channels downstream and upstream. I have never seen a cable modem with the same number of upstream and downstream channels, and I have never seen anything close to even 50% symmetrical on cable internet anywhere in the USA that I have lived. DSL has always had very close to symmetrical numbers. I remember when I was rockin' a full duplex 1.5Mbit/sec DSL line and a cable modem internet line at the same time. I wish I had invested more time in learning about bonding and load balancing at the time so I could've fully utilized both my ISPs
In today’s asymmetric internet environment up and download ratios can be as poor as 1 to 11. While this works well for situations where downloads predominate, it can cause terrible side effects on an otherwise healthy network when an upload saturates the connection, starving the downloads of needed ACK packets.
Ratios worse than 11 to 1 appear to touch closely to the theoretical limit of ~23 to 1 that TCP/IP invokes in it’s control stream.
This is why, in part, worse ratios haven’t appeared in vendor offerings, as downloads are effectively throttled by upload bandwidth.
My speed can test as high at 340/14. I find it pretty cool that my settings allow me to still get higher speeds than the 300/10 that it is rated at while using QoS. I was basically just asking if my settings made sense is all. I've tried openWRT and was using cake QOS it seemed to work just as well, but openWRT is way more complicated than anything I need, I like how DD-WRT GUI is simple, I don't have a complicated configuration. I just want up to date secure firmware and better QoS and other features.
yes, and i told you that the settings usually don't make sense.
Since you seem to have set the value to 98% of your maximum throughput.
I doubt that is sufficient since you usually set 85-90%.
But you skillfully ignored that.
by the way, if you are looking for the highest throughput then do not use QoS
QoS is not for high throughput but for low latency which is much more important.
surfing the internet is much faster with a 100Mbit line with low latency than with a 1Gbit line with high latency.