[Cerowrt-devel] Linksys wrt1900acs rrul traces

Richard Smith smithbone at gmail.com
Thu Apr 7 09:58:43 EDT 2016


A while ago I mentioned that I purchased a Linksys wrt1900acs and I was 
going to do some comparisons of speed/rrul for stock firmware and stuff 
running sqm.

It took a while to happen but I finally got it done.  I was stalled for 
a long time because for a while OpenWRT was not complete for this 
router.  It would boot but you could not (easily) return to the factory 
firmware as updates via the OpenWRT web interface did not work.

Command line updates still worked though periodically I would build dd 
head and see if it got fixed.  Finally one weekend I decided to dig in 
and figure out what was going wrong and discovered that someone else 
beat me to it and it all was working now.

The following flent rrul runs were run here on my local Gbit network 
(pretty quiet) using with OpenWrt Designated Driver r49051 / LuCI Master 
(git-16.081.38806-6b9a743) + Cake installed.

Cake installed via the instructions here:

http://www.bufferbloat.net/projects/codel/wiki/Cake#Installing-CAKE-out-of-tree-on-OpenWrt-rough-instructions

For these tests I had the inbound and outbound limits set to 975000 
kbps.  975000 was somewhat arbitrary.  I wanted it below 1Gbps enough 
that I could be sure it was the router as the limit but yet fast enough 
that I would be able to see the peak transfer rates.

Test setup is:

Laptop<-->wrt1900acs<--->Gbit Switch<--->Netperf server

Prior to running the tests I did trial runs without the wrt1900acs in 
the chain to verify every thing worked at Gbit speeds.

I ran the rrul test for each of the SQM settings at least 5 times.  For 
layer cake and piece of cake each had a run with results that were way 
off from the previous run.  I'm assuming this is some yet to be fixed 
issue and so I ran it 1 more time so that there were 5 runs with similar 
results.

All the raw files are in this tarball:

https://drive.google.com/file/d/0B-P0wCbNmKvAWmRUSzNMdXpUOFE/view?usp=sharing

Hope this is helpful.  I'll be happy to run any additional tests if 
someone wants more.

-- 
Richard A. Smith


More information about the Cerowrt-devel mailing list