Download speeds with extended ethers & Enforce DHCP

[roadhawk]

I have recently been trying to find out where half of my download speed disappeared to after upgrading from 1.6.2 to 1.9.x (Built 20160219-0158 git@3969110).

Here are the results of my experimentation so far:

- "Enforce DHCP assignments" is active.

- /etc/ethers contains 10 lines (i.e. 10 mac / ip pairs).

- Using a linux box, I perform "speedtest-cli" 10 times, averaging about 92 Mbit/s (down) and 31 Mbit/s (up).

- Then, in Gargoyle, I add 20 dummy lines to /etc/ethers, giving a total of 30 lines. Here are the dummy lines:

Code: Select all

ff:fe:fd:fc:fb:fa 192.168.178.140
ff:fe:fd:fc:fb:f9 192.168.178.139
ff:fe:fd:fc:fb:f8 192.168.178.138
ff:fe:fd:fc:fb:f7 192.168.178.137
ff:fe:fd:fc:fb:f6 192.168.178.136
ff:fe:fd:fc:fb:f5 192.168.178.135
ff:fe:fd:fc:fb:f4 192.168.178.134
ff:fe:fd:fc:fb:f3 192.168.178.133
ff:fe:fd:fc:fb:f2 192.168.178.132
ff:fe:fd:fc:fb:f1 192.168.178.131
ff:fe:fd:fc:fb:f0 192.168.178.130
ff:fe:fd:fc:fb:ef 192.168.178.129
ff:fe:fd:fc:fb:ee 192.168.178.128
ff:fe:fd:fc:fb:ed 192.168.178.127
ff:fe:fd:fc:fb:ec 192.168.178.126
ff:fe:fd:fc:fb:eb 192.168.178.125
ff:fe:fd:fc:fb:ea 192.168.178.124
ff:fe:fd:fc:fb:e9 192.168.178.123
ff:fe:fd:fc:fb:e8 192.168.178.122
ff:fe:fd:fc:fb:e7 192.168.178.121

- Reboot Gargoyle.

- "speedtest-cli" again. ONLY 43 Mbit/s (down) and 31 Mbit/s (up). In Gargoyle CPU load for dnsmasq jumps to 22%.

Up to now, that wasn't too weird. Sad, but not weird.
Now the weird bit:

- /etc/init.d/firewall restart

- "speedtest-cli" again. Back to about 92 Mbit/s (down) and 31 Mbit/s (up). CPU load for dnsmasq stays around 3%.

I recently moved from Gargoyle 1.6.2 where this behaviour was not apparent.
I was shocked that 1.9.x cut my download speed in half and that is when I started optimizing.
Now there are just 10 Assigned Static IPs (the kids' devices and a couple of servers) and the speed is stable at 92/31.

I'm on a 100/40 VDSL connection and Gargoyle is on a TP-Link 1043 V2.1

Make of that what you will !

[roadhawk]

[roadhawk]

[Lantis]

I can explain this quite easily.

You're neutering the gargoyle functionality.

/etc/init.d/firewall restart is the incorrect way of restarting the firewall. The gargoyle firewall needs a lot of other things started with it. Doing it this way doesn't start it and hence why you see a performance increase.

The correct command would be something like
/usr/bin/gargoyle/firewall_restart.sh

The reason you're seeing less than top speed is as you stated, the CPU is nearly maxed out.

[roadhawk]

Yes you are right. Thanks for that explanation. I was just beginning to realize that myself after studying the code.

I have deleted two of my posts in order not to mislead people.

[Lantis]

No problem. You aren't the first or last lol.