I am trying to re-establish some baselines today. Normally I would
write a blog post, but with so many new people on the mailing list (we
are at nearly 300 now!)...

So first up was testing upload performance.

flent -H some_server_close_by -t where_you_are --step-size=.05
--socket-stats -l 300 --test-parameter=upload_streams=4 tcp_nup

I am seeing about a 11ms baseline change sometimes. (see zoomed plot
attached) A good supposition might be my path (half moon bay to
fremont) might be ISLd most of the time. Always the worst path has
worse bandwidth. Other explanations desired.

I have temporarily disabled a few flent servers for budgetary reasons.
Still up are:

atlanta,london,newark,sydney,dallas,.starlink.taht.net
flent-fremont.bufferbloat.net
netperf-eu.bufferbloat.net

down went de.starlink.taht.net and fremont.starlink.taht.net, as they
were getting hammered for hours on end, and our bill went up over 150
bucks last month. Whoever that was? please do not hammer via automated
script, these public servers, and set up your own? You learn more by
running your own servers, anyway... or contact me regarding the
bandwidth bill? (I will gladly host servers in any of the clouds we
are in - linode, aws, and equinix) I also took down the servers in
singapore and mumbai as no-one seemed to be using them,
netperf.bufferbloat.net is also down)

But I do encourage everyone to give the public servers a shot a few
times a day, if they want!

-- 
Podcast: https://www.youtube.com/watch?v=bxmoBr4cBKg
Dave Täht CSO, LibreQos