Joined: 28 Sep 2020
|Posted: Wed Dec 29, 2021 2:29 Post subject: Interpreting WIRED performance of routers running dd-wrt ?
|I'm evaluating old (2010-2017) consumer wireless routers running dd-wrt for their ability to act as simple WIRED firewall/gateways for consumer internet.
I'm hoping the experts can comment on my observations. Thanks in advance!
I'm using a client-server setup with a router running dd-wrt in the middle. Client and server run Linux with an iperf3 client and iperf3 server. I also run Apache on the server and use curl as a client to download a single large ISO from the server.
Client Linux laptop wired gigabit ethernet to
--> DD-WRT router LAN gigabit port to
----> DD-WRT router WAN gigabit port to
------> Server laptop with wired gigabit ethernet port running Linux iperf3 server / httpd
My expectation is that I'm testing the software router, the CPU and the wired network hardware. Ignoring radios, ignoring USB, ignoring other services.
Optimal results from iperf3:
|$ iperf3 -c 192.168.20.100 -P1
Connecting to host 192.168.20.100, port 5201
.. some info deleted..
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 1.08 GBytes 928 Mbits/sec 25 sender
[ 5] 0.00-10.01 sec 1.08 GBytes 924 Mbits/sec receiver
Around 937 Mbps is the maximum I've seen and I'm assuming that's gigabit wire speed. With the curl test, I'm looking for 109 MB/s.
My goal with these tests is to evaluate whether a router is unable to route fast enough in these conditions and is a bottleneck to a gigabit connection. Now that I'm lucky to have a symmetrical fiber internet from a telco, I want to ensure that my equipment is not throttling my internet speed. I'll keep using routers that can get at or near above speeds, and give older routers away to people with 40, 100, 300 Mbps internet connections where they continue to work well and don't act as a bottleneck on those speeds.
1) Is this reasonable thinking?
- I know that this is a simple test and it doesn't take into account multiple, simultaneous clients which may represent a heavier load for the router. Also doesn't take into account more complex services running on the router that might be a burden.
- My focus is light internet requests from < 20 devices while maybe 1 download and 1 HD stream are happening.
2) I'm using CTF wherever possible and that is definitely key. From what I've read this impacts the ability to use QOS, but.. not much else?
- My experience is often: Factory firmware is 100% speed, dd-wrt SFE is often less (sometimes 50% less) but dd-wrt with CTF is back to 90-100% of factory firwmare.
3) My tests appear to show that many routers from as far back as 2010 (Netgear WNDR3700v2 from 2010, Netgear R6200v2 from 2013) can pass this test. This makes me want to continue to use them in circumstances where all I need is something that can handle a 40 Mbps down /8 Mbps up apartment connection.
4) For some reason my attempts to run this test on the Netgear R8000 max out at around 850 Mbps despite it having the same BCM4709A0 2 core ARM CPU as the R7000 (which works great!)
- Comments as to why this might be? (Yes, using recent builds from 10/28 to 12/23)
- I'm fishing for what I don't know: If the CPUs are the same, what else is a factor that I can learn more about?
5) My overall observation is that since the primary perceived use of these "Wireless Routers" is the wireless part, non-technical people often blame them for "why my internet is slow" and toss them for something new and shiny. The newer unit's radios do seem *better* (for signal quality) not just faster potential throughput, but for wired performance, I doubt how much better they can be with what I'm seeing the old units do.
- I'm leaving VPN out of this discussion. That seems to be a fairly straightforward case of needing the fastest CPU or dedicated AES hardware. Just walking packet traffic in this thread.
Thanks to those who read this far and those developers (and others) that might want to comment. I hope this helps the community learn, not just me.