HotelOne
Premium
join:2003-06-25
Yosemite National Park, CA
| How Much Speed Loss Due to Wireless Overhead?
I've been unable to get a clear answer to this anywhere. Now that I've got fast DSL with my wireless network I've really noticed the overhead loss on the speed ratings sites as compared to the old dialup connection. I guess the loss percentage is still the same, but the AMOUNT is so much more now, of course. I've tweaked and fooled around changing channels, etc. In comparison to being jacked straight into the modem, I get about a 40% loss of speed when WEP is enabled and only about a 20% loss without WEP. Is a 20% loss due to the wireless overhead about what's expected, and if not what is? Any pointers to web resources would also be appreciated. TIA |
|
vincentfox
join:2003-03-18
Davis, CA
| I really don't see how you can have noticed this JUST because of having gotten "Fast DSL". Most DSL packages I've seen cap out at around 3 Mbps down. A good 802.11b setup will run 3-4 Mbps with good signal.
You don't give much details about your setup, hard to say beyond that. 11b or g? Makers? The better products show little or no speed change with using 128-bit WEP. The crap products, well.... |
|
HotelOne
Premium
join:2003-06-25
Yosemite National Park, CA
| My DSL package isn't capped, and with my lines and distance from the CO I can get about 4 Mbps down straight from the modem to the laptop. Via the wireless network without WEP enabled I get around 3.7 Mbps down, with WEP enabled I get between 2.7 and 3.2 Mbps. My main question is this: Compared to a wired network, about what percent of overhead loss will the average 802.11b show? |
|
ahuebel
join:2003-09-24
Houston, TX
| Well what you can do is download a program like qcheck and run it on 2 computers on your network. Do a test through your wired switch first, and then over the WLAN. On my network I have found net throughput on a 100Mbps switch to be around 90Mbps or so (10%), and on an 802.11b WLAN net throughput is around 4Mbps unencrypted, and on a .11g WLAN, 18 or 19Mbps unencrypted. So if you do the math for the 11Mbps you see about 36% and for 802.11g 33%. These are just what I have found in my network and seem to be about what I have read for others. WEP does provide overhead, again, just do some testing and see for yourself. |
|
HotelOne
Premium
join:2003-06-25
Yosemite National Park, CA | ahuebel--Thanks for your numbers - they seem to be in line with what I'm getting. I'd appreciate input from others here if you could take the time. TIA |
|
briand805
join:2002-10-05
Los Osos, CA
| I am using a linksys befwlls4 v2. that is a 11b set up. 4 computer 2 xboxes connected to the router. Two of the computers are connected wirelessly.
My connection is SBC dsl. My line is rated at 6016 Kbps (downstream)608 Kbps (upstream) I test consistently test at 5000/500 give or take on the wired computers.
ON the wireless conntection the numbers are more like 3500/500 give or take a few. |
|
Angelo_
The Network Guy
Premium
join:2002-06-18 | reply to HotelOne
its different based on where you are you can't say x amount where i use it loss is 0% its just counts |
|
HotelOne
Premium
join:2003-06-25
Yosemite National Park, CA | reply to HotelOne
briand805 - Thanks for the info, my numbers make more sense now. |
|
vincentfox
join:2003-03-18
Davis, CA
| reply to HotelOne
All these radio technologies are inherently half-duplex. So it's better if you compare them to a 10 or 100 connection with both ends set to half, than to your typical modern setup with everything at full duplex. There is significant overhead anytime you have to stop receiving to transmit back acknowledgements and send next packet. Also, just the nature of the radio designs involves much more overhead. 11 megs is never 11 megs, maybe half that. I should toss in that the lower-end wireless AP's will not have enough processing power to do WEP-128 without impact. Later units with an ARM9 chip for example the WEP degradation should be less noticable.
Like someone else said, your best numbers are gotten by doing experiments on local LAN first to isolate the numbers. |
|
ZOverLord
Premium
join:2003-10-20
Minneapolis, MN
edit:
February 17th, @04:57PM
| reply to HotelOne
Actually, when using WEP or WPA, overhead is more based on what CPU is in the card or AP.
With a little research on your card or AP, you can find out what the CPU is in your card or AP.
How Fast is it? How Slow is it? ("Compared to others")
Once encryption is added, the CPU's in your cards and AP's will become the bottle neck of bps.
This can be proven by disabling any encryption and testing rates, and then enabling encryption and testing rates.
It becomes impossible then, because of the differences in these CPU's to make a blanket statement that for example WEP or WPA will cause a speed loss of X%, because it can be smaller or greater depending on the SPEED of the processors involved at the given moment using these algorithms.
Because they are on both ends of the equation, they both effect the outcome. Worse is an AP which has a SLOW CPU, now no matter what CPU's the interface cards have, the AP CPU is SWAMPED with encryption/decryption overhead and slows EVERYBODY down who interfaces with it via a Wireless Device it is hosting.
Other overhead obviously can be attributed to packet retransmission, due to Receiver Sensitivity and Transmitted Signal strength, and Noise, but when compared to the above, in most cases, it's the CPU's that make the biggest CHUNK of lost Bandwidth/Time when using encryption.
You can put 40 foot parabolic antennas on both ends and the CPU's in your Interface cards and AP's are the ONLY constants you can't change, even firmware and HACKED TX power patches can't change that, only a Soldering Iron can 
Bottom line, more focus should be placed on how fast the CPU's in AP's and Interface cards are, because as encryption algorithm's change, and we know they will, how much Bandwidth/Time do you want to give up when you implement them?
As Stated by VincentFox, some of the newer AP's and cards use faster CPU's. It is Interesting that NONE of the name brand manufactures list the SPEED of their processors in their consumer specs.
They may say they use an ARM processor, and you can find out which models run how fast, but they NEVER post the RAW Mhz per second numbers of the processor inside.
Buyer BEWARE, because in some cases this is where the manufactures SKIMP, and sadly, this is the only thing the average person can't change. |
|
HotelOne
Premium
join:2003-06-25
Yosemite National Park, CA | reply to HotelOne
Thanks to everyone for the helpful info... |
|
DrTCP
Yours truly
Premium,ExMod 1999-04
join:1999-11-09
Round Rock, TX
| reply to ZOverLord
said by ZOverLord  :
They may say they use an ARM processor, and you can find out which models run how fast, but they NEVER post the RAW Mhz per second numbers of the processor inside.
I just saw this thread and wanted to add a couple of points.
First Mhz is really a bad indicator. For example AMD Athlon processors outperform the Intel counterparts at the same physical clock speeds. Besides making comparisons acrosss architectures is even more difficult. That is why all those complex benchmarks are developed to have meaningful tests for particular uses.
The CPU speed of the AP may or may not be that relevant at all even for the two AP using the same CPU. It depends on the design. Some AP implement encryption dedicated hardware. You might have a 33Mhz processor but with dedicated hardware doing the encryption/decryption your performance would perhaps equal to 233Mhz processor or maybe even more.
Also there are two types of AP design. One implements most 802.11b MAC functions (including encryption) in the CPU (SoftAP or HostAP) while the other type firmware AP implements most in the wireless module. Obviously, HostAP design is more vulnerable to Host CPU clocking...
There are some newer cards that can do functions partially on the host CPU (most of the MAC) and partially on the wireless module (encryption assist) as a balance of function vs future upgradeability. |
|
