Sudden increase in bufferbloat with Cake QoS when nothing's changed?

[anotherone]

RT-AX86U on Merlin v3004.388.7

Rogers internet advertised speed is 150/30, actually getting around 300-500 down, 30 up. Modem is XB8 in bridge mode.

Using 300/30 as baseline, I set Cake QoS to around 96%, 288.44 down and 28.84 up.

Tested using this setting at 3 am and I got A+ and no increase at all during load. This result was repeated multiple times.

But when I tested later at around 5:45 am, got a B rating and there was a big increase in loaded latency during upload at +56ms.

Why? I didn't change anything between the tests?

I have since rebooted both modem and router, but the increased latency (only on upload) persists.

Cable signals look normal in the modem diagnostic page.

I've been using Rogers cable for ages and upload been at 30Mbps for years and I've been using the Cake setting of 28.84 for a year or two no problem. Only the XB8 modem is new.

Why does this happen?

 

[Morris]

RT-AX86U on Merlin v3004.388.7

Rogers internet advertised speed is 150/30, actually getting around 300-500 down, 30 up. Modem is XB8 in bridge mode.

Using 300/30 as baseline, I set Cake QoS to around 96%, 288.44 down and 28.84 up.

Tested using this setting at 3 am and I got A+ and no increase at all during load. This result was repeated multiple times.

But when I tested later at around 5:45 am, got a B rating and there was a big increase in loaded latency during upload at +56ms.

Why? I didn't change anything between the tests?

I have since rebooted both modem and router, but the increased latency (only on upload) persists.

Cable signals look normal in the modem diagnostic page.

I've been using Rogers cable for ages and upload been at 30Mbps for years and I've been using the Cake setting of 28.84 for a year or two no problem. Only the XB8 modem is new.

Why does this happen?

Your router is not the only one on the internet! There is probably congestion someplace between you and the test server. If all parties between you are the test server are responsible network operators they will see this and fix it.

 

[anotherone]

I ran some bufferbloat tests on another site and it also shows increased latency when upload is saturated.
But Netflix's fast.com test doesn't really show any increase...

 

[bluzfanmr1]

I find that particular website to be inconsistent. I prefer fast.com; click on settings and check each box. The closer the loaded and unloaded latency are, the better.