Answer the question
In order to leave comments, you need to log in
Who cuts the bandwidth of the channel and how to fix it in freebsd+nginx server?
Hello.
I ordered a dedicated server in the USA for sharing video clips + live video broadcasts between users (video chat). By default, 720p is recommended to specify 3500 Kbps Bitrate (youtube, twitch).
We have a 1GB channel, atop currently shows a load of 1%.
The server costs: nginx + Ipfw firewall, SSD disks.
At home, the Internet is 100mb, here is a channel to American servers (I will show one, although many have approximately the same speed):
The problem itself is that when I try to download a file, I get:
I checked the channel to Twitch servers in New York and other American ones, everywhere it pulls a stream of 3500, overclocking is taken to my server and it hardly pulls 1000, 1200 already with drops. nginx-rtmp does not pull above 1000, if you give static like .avi files, then 400-600, it rarely rises to 1300-1400 (nginx).
I asked a friend from another city in the USA to test it, and during the day the reception speed did not rise above 1000 kbps.
Actually, why is this happening? How can I increase the channel and get the coveted 3500-4000 Kbps? Maybe the hoster screwed something up so that the load was not large from 1 connection? Or default settings of nginx cut?
I will add that:
limit_rate
limit_rate_after is 0
by default. I also tried to manually specify 0.
Tests for reading the file gave out 25mb, and the hosters say that nothing is being cut from their side. Our admin also cannot find why the channel is so slow. Somehow, Twitch and other streaming services and dating sites were still able to set up a normal channel. Help, I don't know what to do.
Answer the question
In order to leave comments, you need to log in
Drive tests, for example, raise ftp and try to download them through it, look at the speed or scp / sftp,
but in general, look at nginx
highload
Do the discs work just fine? And then there was a post recently on Habré or GT that some hosting providers are struggling to cut the speed of disk exchange for virtual machines that begin to generate at least some serious load for some time (and not very long).
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question