Suggesting server settings for more pain

[ginger51011]

While working with pandoras_pot, I've found that the raw output power of our hellish pots is never the same in production (duh). This can of course depend on all kinds of things; localhost skips a lot of things that will slow down our pretty pretty packages from dunking on some EC2 instance disrespecting my robots.txt. I just want to raise the question of we should try to figure out possible bottlenecks in a Linux environment?

This excludes everything between you and the recipient; I'm talking "I'm on the same Wi-Fi, I should be getting a lot more out of this". Perhaps I'm measuring wrong?

Suggestions include:

• Actually understanding the Linux network stack
• A program or script for measuring would be nice
• Fiddling with TCP buffer size
• Figuring out how much the S in HTTPS messes with delivery
• Anything smarter than this, like doing some crazy eBPF analyzing (this could be fun!)
• Buy a new router
• Buy a new computer

[yunginnanet]

Best place to start would be tuning parameters in /etc/sysctl.conf. That said, HellPot is cross platform, not sure if you've been testing pandoras pot on other platforms.

I've been planning a realistic measurement apparatus, it's a clunky thing to develop but I agree it's necessary.

Buffer sizes are almost always relevant to performance tuning, without a doubt. Some of these parameters of course exist at the application level, others will be configurable (for linux platforms) via sysctl parameters.

TLS overhead probably isn't all that notable.

//

Use an old recycled computer running linux as your router.

Sell your soul to Satan for a new computer.

[yunginnanet]

also discussions would probably be a better place for this than issues maybe