Achieving 100k connections per second with Elixir
stressgrid.com
Achieving 100k connections per second with Elixir
1–10 of 117 posts
Re: Achieving 100k connections per second with Elixir
#2Would be helpful to know the hardware/instance size they used for these tests. TFA doesn't explicitly state it.
Re: Achieving 100k connections per second with Elixir
#3Would be helpful to know the hardware/instance size they used for these tests. TFA doesn't explicitly state it.
They mention in the Ranch section that "In this test, we set it to 36—the number of CPU cores on our c5.9xlarge."
Re: Achieving 100k connections per second with Elixir
#4Would be helpful to know the hardware/instance size they used for these tests. TFA doesn't explicitly state it.
We used Ubuntu 18.04 with the 4.15.0-1031-aws kernel, with sysctld overrides seen in our /etc/sysctl.d/10-dummy.conf. We used Erlang 21.2.6-1 on a 36-core c5.9xlarge instance.
To run this test, we used Stressgrid with twenty c5.xlarge generators.
Re: Achieving 100k connections per second with Elixir
#5This article is quite good, especially part about bottleneck caused by single supervisor in ranch. However I have to say that title is a bit misleading because all of this has nothing to do with Elixir, it's all about Linux kernel and Erlang, cowboy and ranch are written in Erlang.
Having said that, I will add that I think it is good to have Elixir.
Re: Achieving 100k connections per second with Elixir
#6Would be helpful to know the hardware/instance size they used for these tests. TFA doesn't explicitly state it.
We used Ubuntu 18.04 with the 4.15.0-1031-aws kernel, with sysctld overrides seen in our /etc/sysctl.d/10-dummy.conf. We used Erlang 21.2.6-1 on a 36-core c5.9xlarge instance. To run this test, we used Stressgrid with twenty c5.xlarge generators.
omg.
100K/sec was achieved by yours truly 10 years ago on a contemporary xeon with nothing but nginx and python2.6 - gevent patched to not copy the stack, just switch it. (EDIT: and also a FIFO I/O scheduler)
Why does this require 36 cores today??
Re: Achieving 100k connections per second with Elixir
#7Would be helpful to know the hardware/instance size they used for these tests. TFA doesn't explicitly state it.
We used Ubuntu 18.04 with the 4.15.0-1031-aws kernel, with sysctld overrides seen in our /etc/sysctl.d/10-dummy.conf. We used Erlang 21.2.6-1 on a 36-core c5.9xlarge instance. To run this test, we used Stressgrid with twenty c5.xlarge generators.
skimming fail :(
Re: Achieving 100k connections per second with Elixir
#8Earlier quoted context omitted.
We used Ubuntu 18.04 with the 4.15.0-1031-aws kernel, with sysctld overrides seen in our /etc/sysctl.d/10-dummy.conf. We used Erlang 21.2.6-1 on a 36-core c5.9xlarge instance. To run this test, we used Stressgrid with twenty c5.xlarge generators.
omg. 100K/sec was achieved by yours truly 10 years ago on a contemporary xeon with nothing but nginx and python2.6 - gevent patched to not copy the stack, just switch it. (EDIT: and also a FIFO I/O scheduler) Why does this require 36 cores today??
Was your benchmark for requests/sec or connections/sec?
Re: Achieving 100k connections per second with Elixir
#9Would be helpful to know the hardware/instance size they used for these tests. TFA doesn't explicitly state it.
We used Ubuntu 18.04 with the 4.15.0-1031-aws kernel, with sysctld overrides seen in our /etc/sysctl.d/10-dummy.conf. We used Erlang 21.2.6-1 on a 36-core c5.9xlarge instance. To run this test, we used Stressgrid with twenty c5.xlarge generators.
Did starting more acceptors than the number of cores make any difference?
Re: Achieving 100k connections per second with Elixir
#10Earlier quoted context omitted.
We used Ubuntu 18.04 with the 4.15.0-1031-aws kernel, with sysctld overrides seen in our /etc/sysctl.d/10-dummy.conf. We used Erlang 21.2.6-1 on a 36-core c5.9xlarge instance. To run this test, we used Stressgrid with twenty c5.xlarge generators.
omg. 100K/sec was achieved by yours truly 10 years ago on a contemporary xeon with nothing but nginx and python2.6 - gevent patched to not copy the stack, just switch it. (EDIT: and also a FIFO I/O scheduler) Why does this require 36 cores today??
Duh. Of course a C event loop while be faster at accepting connections, that's not the point of the article.