Any idea for a router transfer speed benchmark ?

David Taylor a =E9crit :

I have a reason but let me explain my problem before :

My ISP provides me a triple-play modem (TV, phone, Internet). The modem is connected to the router. This modem also allows me to watch a picture on TV which is played on my home computer. In my case, I mostly use WiFi to carry the video information.

My problem is the video stream freeze from time to time. If I connect the WRT54GS router to my PC with a long Ethernet cable, the problem does not occur anymore. I also fixed the WRT54GS speed to 12 Mb (instead of letting it auto-negotiating the speed) and I got better result. After desactivating the AES-128 encryption, it was much better. I noticed that the freeze occurs when the stream reach 800 Ko/s.

I think I have a speed limitation problem and that is why I have the freeze.

Consequently, I want to know exactly which maximum WiFi speed I can reach. If I can go over 800 Ko/s, I will be sure the problem is elsewhere.

I have only one PC at home and I do not know anyone to borrow me one. Moreover, I do not want to add one more step in my system since the problem only occurs between the router and the PC.

I thought TFTP was a good idea but I can not reach high transfer rates with that method.

That sounds good. Please tell me if I well understood (because I am French and I do not read English fluently)

Do you mean :

1/ I run a "ttcp" deamon on the PC, another on the router and then run the test to measure the bandwith ?

2/ Or I run 2 "ttcp" on the PC then the packets go to the router using Wifi and get back to the PC using the Ethernet cable ?

What about measuring the speed between a wireless client and a wired client using something like ttcpw? The wired client will more than keep up with the router, so you'll be measuring the wireless bandwidth.

Forgive me for asking but WHY?! :)

The most useful numbers are from the routers ethernet port to another ethernet port or wireless.

I can't think of any reason why you'd want to know how fast data gets

*only* to the router, it generally goes somewhere else like another computer perhaps on the LAN or internet and it's far easier to test the throughput of the router rather than use the router as an endpoint.

So why just the router?

David.

William P. N. Smith a =E9crit :

In facts, I first want to check if I can get a good bandwith. The problem always happen on the same time in the video : when there are many differences from one picture to another. I suppose that only the picture differences are sent to the network. Consequently, a lot of diffs requires a lot of bandwith. If I am bandwith limited, the picture freeze. That is why I am pretty sure this is not a interference problem.

My ISP provides a non-standard modem which runs an home-made firmware. This firmware is download from the ADSL line and they do not give much information about it (the video client is packaged in this firmware which also manages the TV and phone) Anyway, I am going to look in this way.

snipped-for-privacy@laposte.net wrote: [I wrote]

I don't think you can run two ttcpw daemons on one machine, maybe you can with ttcp...

Your real problem, from another followup, is that you are trying to watch streaming video over a wireless connection. Depending on your situation, you may not be able to do this, as interference may cause network degradation.

Can you set your video client to use larger buffers?

The problem there is that with both adapters in the same machine, you're going to be able to route between both internally and never bother going out and back in.

That's why 2 clients, one wired and one wireless with the router in the middle is the normal way of testing the throughput.

You stated that you've limited the router to 12Mbps and that you're seeing 800kb/s throughput, yep that's about right for maximum.

David.

I finally did it !

Thanks to Google, I realized that most of the WRT54G/S firmwares include a version of "ttcp" which is called "epi_ttcp". The binary file is located on the router in "/usr/sbin".

Here my test protocol :

1/ Connect on the router using "ssh"

2/ Forward the 5010 port from the router firmware Web admin page.

3/ Start a SNMP traffic analyser (like PGRT) which directly listens on the router interfaces.

3/ Run the "epi_ttcp" transmitter on the router :

$ /usr/sbin/epi_ttcp -t -v -n100000000000" -p 5010 192.168.1.111

(I used a huge "-n" number to constantly transmit some information /

192.168.1.111 is my PC address)

4/ I ran a SolarWind "ttcp" receiver on port 5010 on the PC :

C:/> ttcpw -r -v -p 5010

I did not need on more PC to do the test and it worked perfectly ! Moreover, I really measure the WiFi bandwitdth (and not Wired + Wireless)

I set my router on bandwith auto-negociation and managed to get a good signal (turning the PC antenna towards the router's antenna) Under XP, I noticed that the WiFi link was synchronized at 54 Mb/s but my test told me that I was transmitting at 1.2 MB/s (9.6 Mb/s) This bandwith is nearly constant (at 9.6 Mb/s) and I have WPA-AES-128 activated.

Do you understand why XP tell me that I work at 54 Mb/s while I really work at 9.6 Mb/s ? Does XP lie ?

William P. N. Smith a =E9crit :

Yes, it is but I think that even if 54 Mb/s is a raw data rate, I should get a better throughput. I read people working at 22 Mb/s and I only get 9.6 !

Ok, in order to check what you say, I will try to do a "ttcp" test with my 10/100 wired Ethernet interface and see if I can reach better speed.

I have no laptop at home. When I get one, I will do the test and compare it to my "ttcp" test.

Because one is your raw RF data rate and the other is your throughput. Isn't this what you are trying to measure anyway?

I still think you'd get better results with a wired computer as the other client, you don' t have any reason to believe that your router has the horsepower to keep your pipe full...

Try it again with a wired connection to your laptop and see what you get. If I get a free minute today I'll try my WAP54G, but don't hold your breath.

This might help. I've posted it about 10 times so far. Time for an FAQ.

This is stolen from an Atheros PDF at:

some additions and corrections by me.

Non-overlapping Modulation Max Max Max Channels ------- | Link TCP UDP | | | | |

802.11b 3 CCK 11 5.9 7.1 802.11g (with 802.11b) 3 OFDM/CCK 54 14.4 19.5 802.11g only 3 OFDM 54 24.4 30.5 802.11g turbo 1 OFDM 108 42.9 54.8 802.11a 13 OFDM 54 24.4 30.5 802.11a turbo 6 OFDM 108 42.9 54.8

The paper claims that encryption is enabled for these calculations, but my numbers seem to indicate that these number are for encryption disabled. Dunno for sure. The Max TCP and Max UDP are the theoretical maximum thruput rates.

SSH is going to exact a heavy toll. I did a quicky comparison of timing an ftp download versus using WinSCP (an SSH-2 client) of a compressed file. My guess is I lost about 1/3 or more performance with SSH. I'll do some more benchmarks when I have time.

Incidentally, thanks for finding epi_ttcp. It's in the Sveasoft Alchemy firmware distribution.

Also, you might wanna spend some time with the "wl" command. It will give you the error rates at various levels and also what connection speed the wireless link is running (per connection). Methinks you'll find that the connection speed varies radically with reflections, interference, and settings. See:

a readable "wl" command summary.

Yeah, I get about 19.3 megabits on a "54 megabit" connection.

Again, I doubt your connection speed is limiting you, I suspect it's you are CPU bound in the router. Trying to run ttcp on your router is very probably your bottleneck.

William,

I ran the test using "epi_ttcp" from the router and got the following speeds :

Using SSH:

Wireless TCP: ~10 Mb/s Wired TCP: ~10 Mb/s

Wireless UDP: ~16 Mb/s Wired UDP: ~22 Mb/s

Using Telnet:

Wireless TCP: ~10 Mb/s Wired TCP: ~10 Mb/s

Wireless UDP: ~17.6 Mb/s Wired UDP: ~22.4 Mb/s

(I got the speeds from the PRTG tool listening on the adapter's SNMP agents built-in the router)

I am still far from the current speeds. A speed 22 Mb/s on an Ethernet 100 Mb/s connection is really too slow !

If I run a top on the router, I can see that "epi_ttcp" takes 90% of the CPU time. So you are right, I think that I am CPU bound to the router so my test is not accurate.

---------------

Jeff,

Thanks for your excellent document about bandwith testing and results and also for the "wl" command. I also think that SSH encryption reduces the bandwith. My SSH speeds are close to my telnet speed so I think it confirms what William said about router CPU limitations.

I have to ask a friend for borrowing me its laptop !

I don't find the XP report very accurate at all, don't know why. Run the software for your wireless card and you'll probably find that jumping all over the place.

David.