No more ICMP drops thanks to SQM

Thanks to everyone who has worked on OpenWrt over the years and to the google employee who ascertained the root of the problem now known as buffer bloat. I have enabled SQM on my OpenWrt on a Raspberry Pi after noticing ICMP packet drops in LuCI monitoring. No more ICMP drops. At first I thought that the graphing was broken, but i was able to confirm that its working, SQM really has solved this problem and i get an A+ buffer bloat rating.

Have a CCNA to my name but still getting my head around the root cause of buffer bloat from a detailed technical perspective. Perhaps I misunderstand the problem because the thought that occurs to me is that if buffer sizes on the internet have increased in a way that is problematic to TCP/IP, have the ITU or other international telecommunications engineering bodies initiated projects to co-ordinate the gradual reduction of these buffers overtime? The problem of IPv4 address depletion was overcome with a systematic plan to phase out IPv4 and phase in IPv6 over a long period of time. Has the problem been addressed at this level or is it too insignificant to warrant the allocation of resources?

After we proved it feasible to make home routers do fq_codel, the IETF formed the AQM working group and published multiple RFCs.

https://tools.ietf.org/wg/aqm/

Jim Gettys, who was working for alcatel, not google, was the father of the bufferbloat movement: https://gettys.wordpress.com/category/bufferbloat/

We consider Stuart Cheshire to be the godfather: http://www.stuartcheshire.org/rants/latency.html

We've given many talks over the years on these subjects. My best was:

Welcome to the party! Please fix it for yourself, then fix it for your friends and family.

2 Likes

Thank you for correcting me Dave and for the link to those resources!

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.