802.11G
Wireless networking continues to advance at a very brisk pace. Keeping up with the standards
and the ups and downs of the new and old products is becoming challenging. The latest
generation of Wireless is called 802.11G which is 54Mb/s. So the standards are:
| Spec | speed | Freq | encryption |
| 802.11B | 11 Mb/s | 2.5GHZ | WEP |
| 802.11A | 54 Mb/s | 5GHZ | WEP |
| 802.11G | 54 Mb/s | 2.5GHZ | WEP/WPA |
WPA is an enhanced encryption algorithm. More about that later ...
More on wireless standards
Also a good intro to wireless standards.
Ok now some basics. As you can see 802.11A is 5GHZ and B/G is 2.5GHZ. So to have a client or access
point to do both means two transcievers meaning it is more expensive. B/G are both 2.5GHZ so only
one transceiver. Where you run into a problem is relative to B/G. Since they are on the same
frequency coexistance is a problem. There are lots of reviews showing some major issues with
coexistance. Tom's hardware site reviewed the linksys WRTG54G
and discussed the coexistance issues in great detail. Tom refers to having 802.11B associated with
802.11G Access points and having them slow down 802.11G. The effects of the slow down can even
continue for some time after the B client has dropped from the G access point. So what is one to do.
Well access points have become so cheap that there is an easy solution. Have one access point for B client and
another for G clients. You can keep them separate by keeping the keys different on the two access points.
That way B clients with never associate with a G access point. I was able to measure a performance hit of
as much as 40% when a B was associated to a G. How long it took to come back to full performance after the
B client was shutdown varied.
802.11G
802.11G is the latest and greatest wireless standard however products came out before the spec was totally
ratified in Sept 2003. So one of the first thing you should do is insure that your firmware and drivers
are all at the latest level to insure the best interoperability. As with early B products I have found
that the vendors has not done a great job of interoperability. As an example I was working with an
IBM 802.11 A/B/G card that uses the Philips chipset (I believe it to be an Accton card) as well as
a DLINK DWL 2000 AP. Similar to what happened with
802.11B DLINK have chosen to allow you to enter a Passphrase while IBM have chosen to allow you to enter a key. The
passphrase is of course vendor Unique. so this means unless you buy your access point and clients from the same place
... it ain't workin! Whenever you see passphrase, pass on the product. Vendor unique.
802.11A
So now what about 802.11A? A acutally came out before G but after B. Only in the computer world. Anyway,
A is 54 Mb/s but does not have the encryption improvements of G. WPA may be added in the future but I do
not see the revenue stream to the companies to justify re doing firmware for the access points and
driver for the clients. Because A works on 5 GHZ it has both pluses and minues. On the same amount of
power A will not travel as far and is more easily obstructed than B/G. What this means is that placement
of the access point is key. You need to do a careful site survey of the home to figure out the best place
to put it. This also means you are more likely to want an access point instead of a router, and you may
need more than one to cover off your home. On the plus side 5GHZ band is less cluttered with things like
wirless phones baby monitors and the like. Also since A was not prolific the chances of having another A
network in your area is less.
Encryption WPA, WEP
Ok so now onto encryption. Since the wireless LAN signals are broadcast they can be picked up by
others. Also with no encryption anyone can come near your home and tap into your network, start stealing
your bandwidth files etc. They could also use your connection to do illegal things. Recently a pervert
was caught stealing interent signals from a wireless connection and using it to download kiddy porn. Don't want
that knock on the door? TURN ENCRYPTION ON. Don't be a moron. You know, I did a drive by using a great
little program called Netsumbler.
This program will find incredible amounts of detail about access points, there encryption level and can
even tie into a GPS to document the location of the access points.
All in all encryption is a must.
Over the last two years there has been much
todo about the weaknesses in encryption provided by WEP. If WEP is all you have available then it is better
than nothing. The issue is that with a brute force attack a person can eaves drop on some amount of
your network traffic and figure out your key. And lets face it, in a home enviroment how often are people
updating there keys? So to fix these problems 802.11i is being worked on. Problem is it is a ways off.
If you want to read a bit about 802.11i this is an excellent article.
A while back Cisco created a technology called LEAP. It was proprietary (only worked with Cisco clients
and access points). This went a long way to solving the WEP weaknesses. What it did was algorthimically
rotate the key. So the WiFi alliance has created WPA based on similar technology. Here's a bit more to read on:
WiFi WPA.
Another excellent article on WPA.
Now there is one thing to keep in mind. WPA is new. VERY new. So this means there may or may not be updated
drivers or firmware for your access point to add WPA. Unfortunately you can only choose one setup. So if you
have 802.11B devices that there is no upgrade for to WPA then you will have to stick with WEP. You can not
have one setting on the access point for 802.11B and one for 802.11G. The only way you could achieve this
is with two access points which as mentioned above is not a bad idea anyway.
In addition to the driver and access point your OS must support TKIP. This is covered quite nicely
on this page from the above link.
There is an excellent quote on this page that says
"In the short-term, if you want to avoid hassles and extra cost with WPA, your OS better be WinXP." Oh ya
and to add some salt to the wound you need a fixpack for XP too.
An excellent article on how to setup a Radius server in a Windows environment
to support secure authenticated wireless networks.
Using more than one access point
Sometimes when you need to cover a large area you will require more than one access point. There is a tip
in the DLINK manual that says you should keep access points that are in proximity to each other at least
4 channels apart to insure that they do not interfere.
Performance measuring
There are a number of things you need to know to if you decide to run benchmarks. Wireless networks can be effected
by spurious noise, other access points, walls, people walking around etc. The net is that the further you are from
access point the higher the variability in your test results.
Test results
Systems used during the test
Server Dual Pentium III 550 on a Intel N440BX running Windows 2003. IDE hard drive.
Client #1 IBM ThinkPad X31 2672c2u with an IBM High Rate A/B/G card based on the Philips chipset. I believe it to be an Accton card.
Client #2 IBM ThinkPad T23 with an Orinoco Gold PCMCIA B card based on the Prism chipset
Mini PCI Vs PCMCIA
PCMCIA is nothing but a new form factor on the old AT bus. This means it is slow. So it adds delays. Mini PCI cards
are the best choice but if you do not have the choice then a CardBus card is the next best choice.
Comparing a PCMCIA card to a Mini PCI card I measured 565 k/s Vs 459 k/s implying that the mini PCI card is 23% faster. This is on B.
I would expect this differenece to increase on A.
Access points
The processor used in access points can have a dramatic impact on performance especially when you turn on encryption. Fortunately most
access points have moved to more than adequate processors. As an example of this I have two access points. An older Linksys WAP 11
and a newer Netgear MR314. I measured 343 Vs 459 implying the newer AP is 34% faster. This can also be due to a better transmitter
and other factors.
802.11B Vs A and G
I did this test with three different Access points and one triple band card
(IBM a/b/g based on Philips chipset I blieve this to be an Accton card). The access points were:
| To | Perf K/s | Notes |
| MR814 B | 560 | None |
| WRTG54G G | 1496 | 267% faster than B |
| DWL 2000 AP G | 1612 | 8% faster than the WRTG54G |
| HW102 A | 1688 | None |
So as you can see there is a slight difference in performance between the Linksys and the Dlink and the
performance on the A is also quite good. These tests were done in close proximity to the access point to
remove variability. As you get away from the access point the A drops off much quicker than the B or G as
previously reported. I did find that with a bit of experimentation I was able to find a good place to put the
A access point to cover my ground floor and get adequate performance.