Kamoj OpenVPN: Easy Speed Optimization up to 115%

[kamoj]

I did this investigation of several reasons.
I always got better results than anyone in the SNB forum.
I have always measured 120 Mbps, others 1/3 of that.
(@R. Gerrits: https://www.snbforums.com/threads/r7800-voxel-internet-speed-seems-capped.67472/post-630758).
(@senpai : R7800 Voxel Internet Speed seems capped ).
https://www.snbforums.com/threads/r...-providers-kamoj-users-are-very-welcome.67907
Why?

Since I started tweaking OpenVPN performance together with Voxel,
a lot has happened to both Firmware, VPN providers and increased
usage of add-ons/plug-ins as Kamoj, Aegis, AdGuard, Netdata etc.
https://www.snbforums.com/threads/c...800-v-1-0-2-40sf-1-0-2-41sf.42222/post-361735

I've always had tweaks that I resist to make public, just because I don't
know how they effect the lifetime of the router.

For this time I just manipulated the Linux and OpenVPN buffers,
and think that can not effect any stability.

The tables here describe the results of just changing buffer sizes.
Note that I only measured download speed, not buffer bloat etc.

The results are very promising, so an upcoming add-on release will have an extra "turbo" gear.

PS
Note e.g. the the very little effect the fantastic Aegis has on performance.
By using correct buffer sizes the impact is only about 5%!
What an achievement by @HELLO_wORLD . Thank you for the amazing program!

The 115% (in Subject Title) is from worst buffer size and a number of add-ons running, to
using few add-ons and big buffers.
PPS
Kamoj add-on 5.4b16 is not public, not even for beta testers -yet.

 

[kamoj]

Reserved for updates

 

[HELLO_wORLD]

Very cool initiative!

PS
Note e.g. the the very little effect the fantastic Aegis has on performance.
By using correct buffer sizes the impact is only about 5%!
What an achievement by @HELLO_wORLD . Thank you for the amazing program!

You are welcome. I try to optimize as far as possible.

Did you get this result with Aegis log enabled or disabled?

Logging is having an impact increasing softirq. This is at the iptables level.
This is generally insignificant, but in my case, someone is letting their large LAN leak into the fiber local collection loop, so I have a lot of requests hitting my router on the WAN side from their local IP to broadcast 255.255.255.255 and iot ports 6666 and 6667; about 5 per second.
When logging, my softirq reaches about 5%, again, low impact on the router, but mesurable, so iptables logging is taking a little toll.

 

[kamoj]

Logging enabled. Should have mentioned that. Sorry!
That means your super cool Aegis takes even less resources!!!

..Did you get this result with Aegis log enabled or disabled?..

 

[HELLO_wORLD]

Logging enabled. Should have mentioned that. Sorry!
That means your super cool Aegis takes even less resources!!!

No reason to be sorry here, you have not done anything wrong.

I like this fine tuning idea, for the VPN, but why not at a wider level for the whole router and playing with other buffers (tcp...)

Next time you run this test, I would be curious to see the difference with log enabled vs disabled.

 

[kamoj]

Thank you for your kindness!
I AM doing other tests, be sure . I never let the router rest...
But my time is limited..
I've added (5.4b16) other tweaking points for tcp, especially for those with Gbit connections.

No reason to be sorry here, you have not done anything wrong.

I like this fine tuning idea, for the VPN, but why not at a wider level for the whole router and playing with other buffers (tcp...)

Next time you run this test, I would be curious to see the difference with log enabled vs disabled.