Author: HQH
Released: March 9, 2004
Updated: N/A
|
Let's begin by looking at DWL-G520's short and long preamble throughput graphs on one plot:
In the above plot, we can see that using 128-bit WEP encryption takes a hit on bandwidth much more than any other encryption
scheme chosen. 128-bit WEP encryption takes the bandwidth down below 40,000 Kbps. Choosing any other encryption (or lack of) puts
bandwidth above 40,000 Kbps. It can also be seen that short preamble allows higher bandwidth, less overhead for CRC (cyclic
redundancy check) error checking/correction than long preamble.
Now it's DWL-G650's short and long preamble throughput graphs one one plot:
In the above plot, there's a few things going on. First, the throughput should theoretically be higher than my desktop
containing DWL-G520 being about 20 feet away from the wireless router. The laptop with DWL-G650 was 2 feet away from the wireless
router, and has a much shorter distance for data to travel to. So this means that there's some hardware limitation going on with my
laptop setup. I would be pointing fingers at my laptop rather than my ftp server computer or the wireless network. My laptop
after all, is about 4 years old. It's barely topping 40,000 Kbps. Secondly, using WPA encryption rapes the heck out of bandwidth on
a DWL-G650. Instead of being 128-bit WEP encryption bandwidth raper that was seen in DWL-G520, it's entirely different with a
DWL-G650. Thirdly, using short or long preamble doesn't really affect bandwidth, well, ignoring lack of encryption for short
preamble. Again, if my laptop can see higher bandwidth throughput from ancient hardware, the short and long preamble would distant
themselves much more just like how the DWL-G520's short versus long preamble bandwidth plot before above's plot. Instead of a few
kilobytes how DWL-G520 did it, DWL-G650 is less than a kilobyte of a difference for each encryption (or none).
Now it's time to see when two wireless clients transfer to each other through the wireless router. Throughput graphs one one
plot below with short and long preambles:
In the plot above, this is bandwidth depicted when two wireless clients are transferring. Here, we see where half-duplex
transmission takes its toll on bandwidth. It cuts available bandwidth in half and then some. If wireless transmission was
full-duplex, we wouldn't see half of the bandwidth missing. This is one of many reasons why anybody would want a higher advertised
wireless product, even if it's probably 1 Gbps (if it ever existed). The higher the bandwidth, the less impact it is when it's
shared. Think all this wireless medium as a hub. A hub shares its bandwidth connection unlike a switch. One moment, only 1 computer
being on could see more than 40,000 Kbps. Add one more user to wireless access, and it's cut in half down below 20,000 Kbps with no
encryption at all! This is one of the downfall of going wireless - if you have a wireless network for humongous collection clients
to use. Bandwidth would dip down real quick. And the plot above only shows two clients hogging bandwidth. Imagine 10 clients on the
wireless router. Each client would only see below 1,000 Kbps when all of them are 100% active. Barely 1 Mbps connection! Blah. That
would royally suck big time. And that's not factoring in any encryption at all. Factor in some encryption, and you'll see way
beyond lower than 1,000 Kbps for 10 clients, probably rock bottom. Also visible to the plot is that whenever having a lot of
wireless clients, it makes much more sense to use short preamble than long preamble to utilize the extra bandwidth. If using long
preamble on a crowded wireless network, you are just asking for a beat down due to deficient bandwidth (unless your wireless network
prone to massive wireless errors).
Below are two graphs consolidating short preambles and long preambles. Sticking both of them all in one plot was asking for
complications, so I had to split up into two plots.
If you're looking to have an over all performance throughput, no encryption with short preamble settings are the way to go. If
you want some security with performance, I would recommend choosing short preamble with WPA (WPA-PSK) encryption. Choosing any
WEP encryption over WPA encryption is like deciding to put up with a short term crackable encryption over a longer term crackable
encryption. Any encryption can be crackable, given that time is a luxury that can be afforded. To give crackers a hard time, you
want to pick the better encryption to have a more secure transmission line. That is why WPA should be recommended. Having long
preamble is a way to detect errors more sufficiently and correcting them. I have yet to really find use of long preamble as it is
more harmful than helpful to me. In terms of bandwidth, long preamble takes a bite out of it (more like a few kilobytes). I would
only advise turning on long preamble if short preamble's bandwidth was actually lower than long preamble. But in my case, my short
preamble bandwidth was higher, so I don't live in a noisy environment that would cause my wireless transmission to produce too
many errors.
However, having short preamble turned on for my router seems to have some type of problem with my wireless router. Within a few
days, the wireless router would start to rack up delay times, and ultimately slow down my wireless bandwidth. I would need to power
cycle (unplug the power cord, replug it) to have my bandwidth and response time back the way it should be. Be on the look out if
this problem surfaces. Power cycling the router should do the trick. However, I've been messing around with my router with a lot of
settings, so it could be my settings making my wireless router unhappy with me. Having the long preamable on however, would take a
week to see problems with the wireless router with the same racked up delay times and slow bandwidth. Again, I believe it's some
obscure and weird settings I be putting into the router that would cause some result like this.
Breaking the 40 Mbps mark is quite remarkable, although not even close to Super G's 108 Mbps theoretical limit. This is why on
average, technology gurus would generally say you'll only see half of what's advertised for the wireless product. This is due to
wireless bandwidth overhead, half-duplex transmission, along with radio interference to impede the overall wireless throughput. We
don't live in an ideal world, so we have to put up with less than expected bandwidth throughput we actually get rather than the
advertised bandwidth we see all over the product's marketing materials. 40 Mbps is less than half, but again, not everybody lives
in an ideal world with ideal conditions. Well, this concludes my findings with D-Link's Super G equipments. I hope you enjoyed the
presentation and find it either amusing, helpful, or down right useless (at least the graphs look sexy right?).
|
|
|
