I sayd what SQM with one interface same result.
Therefore I think , what this problem in OpenWRT packages for shaping.
Get more CPU and you can shape more packets... that's all there is to it. every week people have your same question here on the forum. they want 300MHz single core CPU to handle their new shiny internet connection and it just won't... fortunately you can get a RPi 4 and a switch and a USB 3 Ethernet adapter for about what it costs to have one month of your two shiny internet connections. don't waste your money on two internet connections but a router that is only half as fast as what you need...
Why are you mentioning "two internet connections"?
The OP clearly shows two shapers, and says:
It appears then that he has two WAN connections.
Even if he has only one connection with symmetric 160 Mbps links, this is too much for his CPU obviously. Especially it is too much if he tries to shape both directions. Even with one direction, it is only just barely possible to shape about 150Mbps one direction with this cpu.
I realize that language is a barrier here, but I the conclusion is the same regardless... DDWRT seems to work precisely because it gets about 10% more CPU power due probably to some compilation optimizations.
Shaping is demanding, and what works under ideal testing circumstances and leaves zero excess CPU cycles is not going to work under real world conditions where some WiFi interrupts are required and there are several different connections going at once etc...
If you want to shape anything at all, you should have enough CPU so that under full load with multiple streams you have at least say 20% excess CPU left over to handle all the other stuff that will likely be needed. This device doesn't have that.
I discussed this issue before on a Russian forum and I know some more details.
OP has one internet connection that is PPPoE with IPv4 + IPv6. OP wants to limit speed, no SQM shiny features are needed.
I have no idea why OP made two shapers.
And it is really possible to get ~10% with compilation options. But another problem is that the device has ~6 MB of flash space available.
If this is the case, then the language barrier is even bigger than I thought.
In any case, it's still the case that the CPU handles just about 150Mbps at max throttle. Too little for this link, which can worst case do that in each direction.
The device is too old.
If the goal is to limit speed, then limit speed to something like 100Mbps... this device will do that. SQM in the download direction should handle this fine.
Either accept say 100Mbps shaping performance, or upgrade with something like $100 of equipment and have something that will more than work for at least 5 or 6 years. The network connection probably costs $50/mo right... so get something that will handle it or you are wasting money to get the fast connection in the first place.
It is not the language 
It is lack of knowledge. In Russian it was not easy to get the information either.
1 Like
The OP's problem is that he is using a wierd ISP that provides a 200 Mbps connection, but if you use it for some time at this speed, they cut it to 80. So OP wants to get guaranteed 150 Mbps.
I never heard of this kind of ISP practice before.
Aha, that makes sense what he wants at least.
If this is all he wants, then a simple solution is to buy a managed switch TP-Link sg108 is $27 here in the US on amazon. (hopefully not too different in Russia)
Plug the ISP into port 1 and the DIR-825 into port 2
You can set up in that switch that the egress of the port you connect the router to (port 2) should be bandwidth limited at 150Mbps. This lets the CPU on the device not have to deal with any of the problems 
EDIT: it is also possible to set up ingress policing directly on the DIR-825 and this is probably much less cpu intensive than shaping.... but with the OP's knowledge and the fact that I don't know of any simple point-click policing configuration on OpenWrt... I suspect this is beyond what we can communicate to the OP.
2 Likes
That makes sense. I see the device on Yandex Market for ~$20 with delivery today. So it's not a problem.
Make sure to change the password on the switch!
Would also be good to put the ISP and the router WAN on a separate VLAN that isn't the management vlan... like untagged on VLAN 2 or something.
If OP accepts this solution, I'll try to explain this.
1 Like
i see on TP-Link site that sg108 is an unmanaged switch. Maybe you meant some other model?
tl-sg105e or 108e should fit.
1 Like
yes the 108e or 105e are the managed ones
Which router supported OpenWRT form D-Link 8xx series ? I can buy D-Link DIR-8xx series or other router )
Also , frequancy CPU in DIR-825 not 300 MHz , frequency CPU in DIR-825 680 MHz.
You wanted to say what TP-LINK TL-SG105E need to place in Network with adress 192.168.2.1 ? If so , then I understand )
Switches can have multiple VLANs, that is, separate virtual networks where a packet has a number associated to it and can only go in or out the ports that are associated with that number.... You want to put your ISP device connected to a port which is NOT numbered VLAN 1 because VLAN1 is the place where you can connect to the switch and configure it... that's what I'm saying.
also, YES you want different IP addresses on the different VLANs but the IP subnet and the VLAN are actually separate concepts. you actually can have the same subnet on different VLANs for example even though in your case you shouldnt. You will have to read something a little bit about 802.1q VLANs to make my suggestion work.
Good , I'm can buy TP-LINK TL-SG105E and set so what need to me )
However I want knonw which hardware you can give advice for future.
What hardware can to provide speed least 500 Mbit with traffic shaping QoS or SQM.
Maybe I buy other hardware.
I'm advocating RPI 4 with a switch like that 105e... it can handle 400 or more Mbps with SQM. Anything less is a waste of money I think. Going up from there is an x86 mini PC like Odroid h2 or similar.