Gargoyle Raw Performance

[pbix]

Over the last few days I took some measurements of the raw performance I can achieve on my 200Mhz broadcom router under Gargoyle and a few other software packages.

Downloading speed (WAN -> LAN)
Gargoyle 3600 KiB/sec
CoovaAP (based on whiterussian): 3600 KiB/sec
DD-WRT v24SP2 4300KiB/sec

The upload speed (LAN -> WAN) was lower
Gargoyle 1500 KiB/sec
CoovAP 1400 KiB/sec
DD-WRT 2300 KiB/sec

1 KiB = 1024 bytes/sec

The moral of the story is that if you live in a world with WAN speeds of more than 20Mbps then you need more router to fully utilize.

[DoesItMatter]

What hardware are you running?

What services do you have running when you're doing the testing?

I run both DD-WRT and Gargoyle, and the performance seems
about the same on both.

Only difference I can see is when its running a bunch of services.

Are you doing the testing with QoS disabled on both?

True raw throughput would be no extra services, just the basic
router portion active.

QoS, bandwidth monitoring, quotas, etc all take up cpu cycles and
can show less than expected results.

[pbix]

In the results I shared above QoS was off on all platforms and no special services running. Service would not affect the results anyway unless they consume bandwidth or CPU time.

I also tried disabling the Gargoyle bandwidth monitor by running "iptables -t mangle -F" to remove all chains from the mangle table. The result in Gargoyle was a 200KiB/s increase in throughput. This is about a 5% increase in theoritical throughput. Not significant for most users who do not approach this limit in anycase.

[DoesItMatter]

I just re-flashed my Fonera 2201+

It's what I've been doing all my testing on.

I flashed the latest BS Build 12759 and will do some throughput
testing tomorrow when I get home.

You specifically mention WAN -> LAN throughput - correct?

I'll find some known high bandwidth avaialble locations and
do some large downloads to test true throughput.

I'll probably grab some service packs from Microsoft.Com
I'll make sure its about 100MB or so for a good valid test.

Anything under 5-10MB doesn't give good throughput results.
Especially on my 2 internet service providers, they have a false
speed-boost surge that happens in the first 20MB-ish.

[pbix]

Ok but I used two computers in my house to do this testing.

Using your ISP will most likely affect your results and not show the true performance of the router and give confusing results due to issues like you mentioned.

Instead I recommend you copy a large file from one computer to another. One computer is on the WAN side and one on the LAN side. This is easily done using two routers and two PCs.

BTW, I also did testing using the LAN<->LAN connection. Because of the hardware design of most routers this path does not involve the software. I clocked this speed at 7600KiB/sec.

Good luck

[DoesItMatter]

OK... did some testing

I have the following setup configured and this is how the
throughput tests were done with both firmwares.

Cable Modem -> Router 1 WAN port (main router)

Router 1 LAN port -> PC 1 NIC
and
Router 1 LAN Port -> Fonera 2201+ WAN

Fonera 2201+ LAN -> PC 2 NIC

PC 2 NIC is setup to be DMZ on Fonera 2201+

First I flashed latest DD-WRT build 12759 on the Fonera 2201+
Did a throughput test from PC 2 -> PC 1 with 250MB file.

Next I completely wiped and flashed Gargoyle 1.0.8 on the
Fonera 2201+ and throughput test PC 2 -> PC 1 with 250 MB file.

Attached are the 2 test results.
No QoS or Quotas or anything setup on either of the 2 routers.

[Eric]

Thanks guys, for running these performance tests! I suspect that performance is very hardware specific (hence different numbers for Fonera vs DD-WRT), though I'm not entirely surprised DD-WRT is a bit faster, since it's more related to the original (optimized) linksys firmware than the latest versions of OpenWrt. I suspect you'll get almost identical performance statistics for a default install of OpenWrt as you're seeing for Gargoyle (which, if you have time, would be another interesting test).

I'm especially glad to hear that my bandwidth monitor/quota system is causing at most a 5% performance hit. (Of course, I designed it to be that way )

[pbix]

Interesting results DoesItMatter. It seems your Fonera box is very much slower than the WRT54G.

But it also is interesting to me that your upload / download speeds are the same. There is a significant difference on my setup.

Eric,
Any ideas why the bandwidth is not asymmetric?

[DoesItMatter]

Grab the LanSpeedTest file from the website listed in the images and see if you can do some comparable tests to compare the results.

If I understand your results, from your first post:

Downloading speed (WAN -> LAN)
Gargoyle 3600 KiB/sec
CoovaAP (based on whiterussian): 3600 KiB/sec
DD-WRT v24SP2 4300KiB/sec

The upload speed (LAN -> WAN) was lower
Gargoyle 1500 KiB/sec
CoovAP 1400 KiB/sec
DD-WRT 2300 KiB/sec

1 KiB = 1024 bytes/sec

On mine, I'm looking @

about 1400 KiB/sec up/down on the Fonera with Gargoyle 1.0.8
about 2700 KiB/sec up and 2250 KiB/sec down
with DD-WRT latest version for the Fonera

So... not sure, some parts slower than a Broadcom device,
but other parts are faster?
Yeah - No idea why my up/down were so matched in speed.

[DoesItMatter]

I re-did the tests this AM when I got home and made sure any
services I don't use were disabled.

Same test setup as above, with 250MB file.

Performance increased about 2.5 Mb/s each way up/down