https://netdevconf.info/0x15/session.html?keynote-balakrishnan
Hari Balakrishnan did a 1hr keynote on the "bundler-project", which aims
basically to generate a estimate of available bandwidth between sites.
i.e. between your corporate network or DC, and some other DC (e.g. EC2
us-west-1)
The goal being that, if one can estimate the available bandwidth, then one
can move your congestion from some unknown cloud router to one's edge router
where one can better control it with, e.g. fq_codel.
Of course, this only works for traffic to the places where you've deployed
bundler, and there is the problem of other people's traffic also going to
that destination ("cross-traffic" he called it).
I found it interesting.
It doesn't solve the problem that I've long been obsessed about, which is
finding a way to estimate the total bandwidth cross that "l2" hop between
a home CPE and the ISP "CMTS" or "BNG"... such that we can auto-tune things.
IETF watchers might also care about the APN (really poorly named in the end)
BOF, where one use case they care about is where the "L2" hop is actually a
large metro-ethernet.
--
] Never tell me the odds! | ipv6 mesh networks [
] Michael Richardson, Sandelman Software Works | IoT architect [
] [email protected] http://www.sandelman.ca/ | ruby on rails [
_______________________________________________
Codel mailing list
[email protected]
https://lists.bufferbloat.net/listinfo/codel
