Joined: 12 Mar 2018 Posts: 57 Location: Mont-St-Hilaire, Qc, Canada
Posted: Sat Dec 05, 2020 16:57 Post subject: Are bufferbloats related to router inherent performance?
I have a 100/30 cable connection with Videotron (if that maters).
I have an Archer C7v2.
If, while I am the only one active on my network, I do repetitive DSLreports tests, I get an average of 115/35 but the bufferbloats are ranged from C to D and the quality are also ranged at D. Even with 100/30 connection, those bufferbloats really show on simple web browsing with much noticeable latency in page rendering.
If I enable Qos per @MSOENGINEER settings, as one would expect, I lose some speed and get an average of about 90/20 but bufferbloats gets to A+ or A and overall quality to A.
But then again, I do this while there is no one else active on my network.
My stepbrother has the very same plan with Videotron (he lives in another town so not same server), but when he runs DSLreports on an unknown router w/o QoS, he gets both fast speed and low bufferbloats. Damn!
How is this ?
Is it because an Archer C7 is so cheap that it cannot handle a single client w/o any packet management enabled or is there anything else that eludes me?
Should I investigate something ?
Thanks for any enlightening in that mater. _________________ [Supermicro X10SBA] pfSense as of 20/12/20 | firewall, gateway, routing, QoS, adblocking
[Archer C7(CA)v2] running r47510 as of 21/10/06 | WAP
[Archer C7(US)v2] running r47510 as of 21/10/04 | napping backup for Gateway, routing, QoS
buffer bloat is a function of the amount of buffer set for a stream. The internet was designed to have systems will small buffers and be able to re-transmit. As memory has gotten so cheap, it became easier to add more memory and increase the buffer sizes. This is what is detrimental. Instead of the built in tcp congestion control or other methods such as qos being able to do their jobs, the buffers just fill up on the routers, and they will get emptied but the rate of processing.
If there is a high processing rate, the buffers should not get full or stay full for long, but at a lower processing rate they stay filled longer.
It is not really a sign of cheap router performance but a cross between settings on the system and the processor to memory (or buffer) ratio.
A slow processor with a small amount of buffer works just as well on buffer bloat as a fast processor with a larger amount of buffer. THIS IS CONSIDERING THE BUFFER BLOAT ONLY. other things will be different.
Often times cheaper systems are not engineered to balance these types of considerations well. For instance a faster processor with not much memory has its problems, but a faster processor with TOO much memory takes a long time as well.
Joined: 12 Mar 2018 Posts: 57 Location: Mont-St-Hilaire, Qc, Canada
Posted: Tue Dec 08, 2020 12:56 Post subject:
Thanks both _________________ [Supermicro X10SBA] pfSense as of 20/12/20 | firewall, gateway, routing, QoS, adblocking
[Archer C7(CA)v2] running r47510 as of 21/10/06 | WAP
[Archer C7(US)v2] running r47510 as of 21/10/04 | napping backup for Gateway, routing, QoS