I see that a lot of folk here are now aware of bufferbloat, fq_codel, sch_cake and smart queue management (SQM). Universally (aside from running out of cpu), the reports are positive, which makes me happy (and happier still that y'all seem to "get" it).
What I'm interested in finding out though, are what are the drop/marks per day people are seeing in their deployments?
Recently I started calling these "debloats/day", because I know that every drop is preventing a 500+ms latency excursion on my campus network (3000/day inbound on just the main gateway, 12,000 on my wifi networks, 100+ outbound), and I feel good every time a packet is lost in this cause.
How many drops/marks a day are y'all getting with normal traffic?
(I monitor these with mrtg personally)
In linux the command to see this is "tc -s qdisc show dev yourdevice"
No comments:
Post a Comment