Discussion of maximum throughput of a SOHO router

Post new topic   Reply to topic    DD-WRT Forum Forum Index -> Broadcom SoC based Hardware
Author Message
linksyshackn
DD-WRT Novice


Joined: 19 Sep 2006
Posts: 12

PostPosted: Tue Sep 19, 2006 23:29    Post subject: Discussion of maximum throughput of a SOHO router Reply with quote
Hi there

can others please confirm whether they have similar slowdowns with their SOHO routers regarding WAN to LAN speeds?

I have ran all permutations of:

- WRT54GSv1.0
- Buffalo WRT-WHG-G54S
- dd-wrt-c23 SP2
- dd-wrt-c23 SP1
- each vendor's latest specific firmware

and the results are conclusive:

With SOHO router:
between 2.4Mbps and 4.4Mbps down

Without SOHO router (direct cable modem connect):
between 6.5Mbps and 8.1Mbps down

I had heard that SOHO routers were underpowered, but I just didn't realize it would have this much impact.

I am interested in your real world results.

Where is the slowdown coming from?

It does not appear to be CPU bound nor RAM bound judging from the stat's in DD-WRT. One of these SOHO routers has 32MB RAM.

I know low cost NIC's can cause a problem. But am I to believe that the built in NICs are the sole cause of this problem? All hardware NIC bound and the CPU/RAM is just sitting idly by waiting?

Are there any SOHO routers that can achieve 9Mbps downspeed?

Has anyone any experience on the throughput of the following:

- clarkconnect
- pfsense
- m0n0wall
- ipcop

And would throughput improve if dd-wrt was run on a WRAP Soekris (can it?)?

Note, this is NOT a P2P timeout problem. No P2P activity here, just a simple raw bandwidth test.

I would be interested in your comparative results from here. Try different servers, sometimes the closest one is not the fastest:

http://www.speakeasy.net/speedtest/
Sponsor
dicksons
DD-WRT Guru


Joined: 07 Jun 2006
Posts: 980
Location: Coal Creek Canyon, Colorado

PostPosted: Wed Sep 20, 2006 1:34    Post subject: Reply with quote
if you're trying to isolate variables, runing across the internet to speakeasy sites is probly the best bad way to do it.

run some tests from one machine on a lan or wlan side to another directly on the wan port and use that as a yardstick.

compare that to lan-lan or wlan-wlan or wlan-lan -- that would be a valid way to compare different brands/models of soho router.

if you tested all that for various different dd-wrt compatible routers, we'd all love to see the results !

_________________
linksys GSv2, Gv4, Gv2, GLv1, G-TM, Buffalo wbr2, whr, whr-hp, whr-g125, wli-tx4-g54hp, Moto wr850gp, Alix.3C2
linksyshackn
DD-WRT Novice


Joined: 19 Sep 2006
Posts: 12

PostPosted: Wed Sep 20, 2006 3:30    Post subject: Reply with quote
Thanks...I totally agree setting up a controlled server is preferable...

but unfortunately, it doesn't change the fact...

my results were all too repeatable and not due to variations in speakeasy's bandwidth, as I took multiple readings and had dramatic differences within seconds of switching back and forth, which I did multiple times.

But I agree, it would be useful to set up some ftp or http servers on the WAN side and test to help isolate some more. Perhaps someone has already done this as at the moment I am short of machines but I will do this later...

I am curious if anyone is getting any more than about 4500kbps from the WAN? And why might this be?
linksyshackn
DD-WRT Novice


Joined: 19 Sep 2006
Posts: 12

PostPosted: Wed Sep 20, 2006 5:57    Post subject: Reply with quote
heres an interesting thread discussing the relative poor performance of the soekris board WAN to LAN and possible reasons why they are not seeing the 17Mbps they are supposed to see on a net4501.

they theorize it may have to do with interrupts...

http://www.nabble.com/Net-4801-performance-t2211328.html
PsyDuck
DD-WRT Novice


Joined: 05 Jul 2006
Posts: 10
Location: Sthlm, Sweden

PostPosted: Wed Sep 20, 2006 8:31    Post subject: Reply with quote
Are we talking Mbit (or do you mean Mbyte)?

In my setup I have:

Internet <-> WRT54G v2 <-> WDS <-> WRT54G v2 <-> Client PC

I have no trouble maintaining a stable throughput of 9Mbps in either direction.

I've tested the linksys with stock firmware when I first got it (about 2 years ago) and remember seeing something like 30Mbps throughput WAN->LAN.

I'm not doing any QOS and the firewall is only using default settings.
nlinux
DD-WRT User


Joined: 07 Jun 2006
Posts: 70

PostPosted: Wed Sep 20, 2006 13:11    Post subject: Reply with quote
I would suggest getting a throughput test suite and running it in a known environment. Relying on internet tests for anything other than ballpark estimates is just not going to give true performance figures. But to satisfy your curiosity, I ran the speakeasy.net test and I am getting approximatley 10 to 11 mbs throughput both with and without the router.
_________________
*********************
DD-WRT v23SP2
WRT54GS 2.1 @ 216 Mhz
SD Mod w/ 512 MB Card & Dual Serial Ports
rkramer
DD-WRT User


Joined: 07 Jun 2006
Posts: 71

PostPosted: Wed Sep 20, 2006 18:33    Post subject: Reply with quote
tests on the open internet are useless and should be disregarded... WAAAAAAY too many factors. how do you know your ISP isn't discriminating against certain mac's? it happens, some notice abuse from a certain brand and throttle them down. how about MTU? having the MTU off by only 1 byte can kill your speed.

as someone else mentioned, just get two computers, one with iperf server and one as an iperf client, and route through the router on your local network. REALLY easy to do and will give you the true absolute bandwidth.
linksyshackn
DD-WRT Novice


Joined: 19 Sep 2006
Posts: 12

PostPosted: Thu Sep 21, 2006 0:37    Post subject: Reply with quote
PsyDuck wrote:
Are we talking Mbit (or do you mean Mbyte)?


i meant Megabit. 4.5 Megabit per sec is my current WAN to LAN. If I pull the router out of the loop it jumps to 8 Megabits

PsyDuck wrote:

I have no trouble maintaining a stable throughput of 9Mbps in either direction.

I've tested the linksys with stock firmware when I first got it (about 2 years ago) and remember seeing something like 30Mbps throughput WAN->LAN.

I'm not doing any QOS and the firewall is only using default settings.


thats good to know and quite impressive.... I've turned off the SPI firewall and I'll make sure my QOS is off and we'll see.
linksyshackn
DD-WRT Novice


Joined: 19 Sep 2006
Posts: 12

PostPosted: Thu Sep 21, 2006 0:39    Post subject: Reply with quote
nlinux wrote:
I would suggest getting a throughput test suite and running it in a known environment. Relying on internet tests for anything other than ballpark estimates is just not going to give true performance figures. But to satisfy your curiosity, I ran the speakeasy.net test and I am getting approximatley 10 to 11 mbs throughput both with and without the router.


great, thank you... and i agree i'll do a better test by next week when i'm more equipped.

in the meanwhile its good to know that there is something wrong with my setup that i'll recheck...
linksyshackn
DD-WRT Novice


Joined: 19 Sep 2006
Posts: 12

PostPosted: Thu Sep 21, 2006 0:46    Post subject: Reply with quote
rkramer wrote:
tests on the open internet are useless and should be disregarded... WAAAAAAY too many factors. how do you know your ISP isn't discriminating against certain mac's? it happens, some notice abuse from a certain brand and throttle them down. how about MTU? having the MTU off by only 1 byte can kill your speed.

as someone else mentioned, just get two computers, one with iperf server and one as an iperf client, and route through the router on your local network. REALLY easy to do and will give you the true absolute bandwidth.


interesting, i never thought of MAC discrimination or knew that cable co's do that. now, in this case, the cable modem is the endpoint MAC, regardless of whether i plug the computer into the cable modem or the router into the cable modem (not doing any MAC cloning), so i don't think that's it.

MTU - now that is interesting.... last week i did a test where i lowered the MTU on the WRT54GS to an arbitrary lower number (maybe 1420) in an attempt to avoid any fragmented packets.... but didnt snoop the packets or anything to see what was really happening and i am not an expert so it was just a shot in the dark and it unsurprisingly didnt fix anything.

I didn't know about iperf, i just checked the man page and it seems pretty cool. i'll definately try that when i get some of my computers back online...

but its good to here these high WAN to LAN speeds, i didn't know that was possible on a WRT54G that makes me feel better.

i'll update here what happens..
linksyshackn
DD-WRT Novice


Joined: 19 Sep 2006
Posts: 12

PostPosted: Fri Sep 22, 2006 17:40    Post subject: Reply with quote
Thanks for the iperf tip. That is one sweet app. So much simpler than sending FTP/HTTP files about.

Well you guys were absolutely right.

iperf revealed about 21Mbps WAN to LAN.

so in my mind that completely rules out the question of the SOHO router being underpowered for today's broadband speeds...

so what does that mean? its basically a client server sizing mismatch causing the almost halving in broadband speeds to many web sites / ftp sites during large file xfer when the router is in the middle?

I am not an expert so can someone help me define an appropriate testing strategy, perhaps modifying the TCP window or MTU or what not, and I'll report back the results here?

Thanks!
Display posts from previous:    Page 1 of 1
Post new topic   Reply to topic    DD-WRT Forum Forum Index -> Broadcom SoC based Hardware All times are GMT

Navigation

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You can attach files in this forum
You can download files in this forum