Wireless to Wireless Performance

[bcronin]

I have 3 computers that all have Dlink G810 wireless bridges attached to their ethernet ports. They all talk to a Dlink G800AP access point that is connected to a Netgear 10/100 switch (where my Dlink 604 cable router is also connected). There's two more computers connected to the switch using cat5.

Here's the problem.

Copying large files from one of the wireless computers to one of the wired computers works great, very fast. However, copying large files between two wireless computers crawls (kbits/second as opposed to mbits/second with wireless-to-wired).

Now, I realize that in order to go from wireless to wireless, the data has to flow to the access point and then back to the other wireless computer, so I would expect less performance, but not from mbits/sec to kbits/sec!

What's going on here? Any clues on things I could try to fix it?

Thank you in advance for any assistance you might be able to provide.

BC

 

[kwebb]

You might play with the MTU size in the registry. There are Graphical tools as well that will allow you to adjust MTU size. www.speedguide.net will have one or two. Other than that, firmware upgrades if available. What are you pings going across one of the bridges?

 

[bcronin]

MTU is 1500 on all computers, standard for ethernet I think. Firmware is current on all brides and the access point. Pings across the bridges are fine, typically 1ms (although I did try some pings with longer length packets, e.g. ping -l 4096 and -l 16384) and once it got "too big" things did start to go to heck (but I don't really know what I am doing ;-).

 

[kwebb]

Right, you have the default MTU settings. Kinda figured that. My suggestion is to play around with your MTU settings to see if you see a difference. You might also download and run TCPDoctor from speedguide and see what it suggests.

 

[goloap]

You should expect approximatly half the speed you had when you did wireless to wired transfer. Even a bit less then that. But you seem to say that it is way more than that. Check the settings for fragmentation treashold and RTS/CTS threashold. They should be as high as possible if you are in a good wireless environment (i.e. not to much interference).

Also the MTU is a possible problem, but I think it is less likely since it seems to work well when going to the wired network.

In ancient times they had no statistics so they had to fall back on lies

 

[kwebb]

Nonsense. Well let me step back. Wireless as a technology won't cause your speed to be any different in a particular direction when you go wired/wireless or vice versa. Perhaps the budget AP/clients have this characteristics. The enterprise gear certainly doesn't and my SPeedstream SOHO router/ap combo also does not. I'd be interested in other comments on this. Do you really see significant difference in througput depending on the direction of your transfer? I don't know if I could live with that. Think I'd have to return the gear. It certainly is not true as a blanket statement but my experience is predominately with corporate equipment and the odd budget hardware. 3Com's AP's, can't remember the model don't exhibit this behavior and Siemens either. I've used other cheap stuff but not really enough to notice that or do any benches.

 

[goloap]

I'm not saying that your upload speed should be different than your download speed. I'm saying that if your transfer a file from one station to the other station then the transfer speed will be half of the tranfer speed of one station to the network. Here is an example:

STA1 and STA2 are connected to the same AP:

STA1 ---- AP ---- Network
STA2 ____/

If you transfer a file from STA1 to Network (in ideal conditions) the transfer speed will be at the maximum (lets say for .11b 6Mbps). Now, if you keep the conditions the same, but decide to transfer a file from STA1 to STA2 then the transfer speed will be approximatly half of what you got before: i.e. 3Mbps. This is because you have the data has to pass twice on the air interface (once downlink and once uplink) in contrast the first scenario the data only past once on the air interface (only uplink).



In ancient times they had no statistics so they had to fall back on lies