This was an example for the types of VPNs commonly used.
Please read the second post for more clarification: https://news.ycombinator.com/item?id=12236645
ChaCha20 - not because it has hardware acceleration, but because it's fast in spite of not having it
So unless you're using OpenVPN it's not going to 'crush' an RPi4 then...
Surprisingly, there are lot of users still using openvpn. And armv8 crypto extensions isn't just used for OpenVPN.
It's clear that cortexa72 cores will beat the cortexa53 cores easily in terms of raw performance.
Raspberry PI is designed in such a way that it can be applicable in lots of other use-cases. ML,AI, Image Processing, Desktop, etc, IoT. But, in reality, considering a network application, these features are useless as the openwrt system will be running headless.
Imagine buying a Nvidia Jetson Nano just to run OpenWrt on it.
Raspberry Pi 4 has been a choice for everybody because of its easy availability and the huge community support.
So we're agreed that your claim was wrong then..
Let's not forget the price, which as far as I can tell is in the same ballpark as the NanoPi R2s.
Aren't we discussing about armv8 crypto and OpenVPN and not raw performance of cores?
No, we're discussing your claim that as a 'VPN' router (shadowsocks/wireguard/openvpn) it'll crush the RPi4. It will only do that if you're using openvpn. If you're using either of the other two VPNs you mentioned then it won't be the case. It will also potentially stop being the case with OpenVPN later this year when they release OpenVPN 2.5 with ChaCha20-Poly1305 support for the data channel.
Yeah, it would surely.
Shadowsocks has a variety of crypto algorithm, AES included. OpenVPN is also AES.
That leaves us with Wireguard, which uses chacha20, and thats actually really good.
But, out of the most commonly used three named above, only two supports TCP. Thats shadowsocks and OpenVPN.
Its also been known that TCP over TCP is a bad idea. But with congestion algorithms like bbr, it isn't true anymore. Most ISP and organisation throttles and even blocks UDP packets. My ISP is known to throttle UDP packets, even on port UDP/53. For users in China, this is more common. Also MPTCP works over TCP, thats another reason.
the build failed:
make defconfig
fatal: Invalid revision range ee53a240ac902dc83209008a2671e7fdcf55957a..HEAD
fatal: Invalid revision range ee53a240ac902dc83209008a2671e7fdcf55957a..f39366626c335ce6e8cb1f7c6a2c220e772d8b83
fatal: Invalid revision range ee53a240ac902dc83209008a2671e7fdcf55957a..HEAD
fatal: Invalid revision range ee53a240ac902dc83209008a2671e7fdcf55957a..f39366626c335ce6e8cb1f7c6a2c220e772d8b83
WARNING: Makefile 'package/feeds/packages/sqm-scripts/Makefile' has a dependency on 'kmod-sched-cake-virtual', which does not exist
#
# configuration written to .config
#
do you have the same warning ?
Warnings can be ignored.
thanks i will try again using a fresh ubuntu container.
edit: still build failed
i think we need python-dev and swig on the host machine (see the other thread for the pi2rs)
Thanks for all !
succes build under wsl2 ubuntu 20.04 on windows 10
here the process:
sudo apt-get update && sudo apt-get upgrade
sudo apt-get install build-essential libncurses5-dev libncursesw5-dev zlib1g-dev gawk git gettext libssl-dev xsltproc wget unzip python
# add the following dependecy otherwise the build fail (not sure wich python version dev package is needed)
sudo apt-get install python3-dev swig
cd ~
mkdir -p /openwrt/source
cd /openwrt/source
git clone --depth 10 https://git.openwrt.org/openwrt/openwrt.git
git clone --depth 10 -b gpio https://github.com/jayanta525/openwrt-nanopi-r2s.git
cd openwrt-nanopi-r2s
cp -r ../openwrt/package/kernel/kmod-sched-cake-oot package/kernel/kmod-sched-cake-oot
./scripts/feeds update -a
./scripts/feeds install -a
time make -j$(nproc) download V=s
make defconfig
# remove all windows10 environement variable otherwise the build fail
PATH='/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games'
time make V=s VERBOSE=1 -j$(nproc) JOBS=$(nproc) 2>&1 | tee build.log | grep -i -E "^make.*(error|[12345]...Entering dir)"
In Europe, better buy from antratek.
python2-dev and python-dev has to be removed, otherwise you will have errors at compilation time. The full list would be:
build-essential libncurses5-dev libncursesw5-dev zlib1g-dev gawk git gettext libssl-dev xsltproc wget unzip python python3-dev swig
I've got the nanopi for a week and it's been really stable, no crashes or something strange. I think the build its ready for PR.
I found that the nanopi its really hot, i've got the metal case and somehow its okey, its at 40º at idle, but with SQM and other things the nanopi reachs 60º with just a 40% cpu usage... for my real world use its fine, but if one day i upgrade my conecction i dont know what temps its going to reach.
Sqm work even with the following warning " WARNING: Makefile 'package/feeds/packages/sqm-scripts/Makefile' has a dependency on 'kmod-sched-cake-virtual', which doe
s not exist" ?
You are using wich metal case ?
No, you have to download the latest master and replace the Makefile from package/kernel/kmod-sched-cake-oot to the nanopi source, just after the git clone command.
This one
thanks
i have updated the step by step by step guide.
i have found a new "lowcost" metal case
here:
https://www.banggood.com/All-Metal-Protective-Case-for-NanoPi-R2S-Open-Source-ARM-Board-DIY-Part-p-1686994.html?rmmds=search&ID=6287830&cur_warehouse=CN
And here:
https://www.fasttech.com/search?Metal%20Protective%20Case%20for%20NanoPi%20R2S
Yes, i using without fan metal.
Those two metal cases are different.
One is with a spot for a fan
One is without fan.
Which one are you using?
I using without fan metal.