Slashdot Mirror
802.11n: High Throughput, Not Just Fast Wireless
eggboard writes "Unstrung reveals that the 802.11 working group is spawning 802.11n, a high-throughput task group to work on increasing the actual data:symbol ratio in wireless networks while also boosting speed to 108 Mbps to 320 Mbps. Most people who use 802.11a, b, or g know that actual net throughput, or the real data that's carried, is a fraction of the cited rate: maybe 7 Mbps in the 11 Mbps 802.11b flavor and 25 Mbps in the 54 Mbps a and g flavors. The goal of 802.11n is to increase speed, sure, but also to increase the percentage of symbols that don't bear overhead. The bad news: they predict 2005 or 2006 for completion."
Come on - by then UWB coupled with cellular and mesh will offer those speeds. "n" should stand for "n" significant.
I think I speak for everyone when I say put egos aside, gather the best of each protocol into one protocol and make it the standard and release products for it. I think people (in general) are scared right now because they don't want to buy a product that will not be usuable in a year or two. I want my WiFi already.
Visit www.seriouslythough.com
so why hasn't this been done before? Shouldn't a good spec already be lean? We've gone through six incarnations or so of 802.11 and we still have this kludge going around? I'm sure its not as easy as I'm making it out to be, but I would get fired if someone could take my spec and cut out this much overhead from it. The time to spend the four or five years crafting a good symbol set is before the thing goes public, not after.
If you want highspeed wireless networking just buy some of the at&t microwave bunkers. As a bonus you can have huge w/lan parties and microwave your food free of charge at the same time.
Checking out my form of escapism.
Favorite quote from the article:
802.11 Specification letter suffix: O
What it does: Not being used, because it looks confusing.
In all seriousness, this would be an incredibly useful technology--802.11b at it's current real speed is quite unusable for transferring files of significant size. However, I have to admit that I'm tired of seeing a Baskin-Robbins offering of wireless flavors...802.11g is a noble effort at standardization, but backwards-compatible technology is a must. I can't afford to have twenty 802.11g, 802.11n, 802.11-whatever cards sticking out the side of my notebook/PDA/random useless all-in-one wireless device of the week.
I'm glad someone is focusing on the overhead and efficiency of the standards and not just trying to get something out there with a big unrealistic marketable speed. I guess comparing advertised Mbps on wireless devices could be like comparing MHz for CPUs by different companies.
This has to do with information theory, source and channel coding and modulation.
It'd be nice if these people standardised on a framework that can be combined with various coding and modulation schemes, in a modular sense, instead of creating 802.11xyz groups every now and then...
Guess marketers and managers (ie The Incompetent at best, The Illiterate as per usual) have taken over from the engineers.
/. Where the truth
No real details on what they're trying to achieve.
Are they aiming on improving software, the hardware or both? Anybody know more about this?How small a thought it takes to fill a whole life
The speeds quoted are burst speeds. Assuming
binomial distribution of arriving packets (and that is the regular assumption), the throughput
for any packet based transmission is ~60% of the
burst speed. So 7MBPS out 11 is right what you
should expect.
Have they taken into consideration that they might run out of letters at some point? ..especially if we skip from g to n.. there's some cool letters in there...
Throughput on 802.11b networks is actually only around 3-4 Mbps in optimal conditions, less than 50% of the theoretical maximum.
Too many workstations on the network and you got a recipe for collision. There are just too many factors to dramatically increase throughput.
-----
One is born into aristocracy, but mediocrity can only be achieved through hard work.
They might as well just give up and start saying "Oh, you don't want 802.11[n] anymore - you should throw away all your hardware and get 802.11[n+1] instead, since it'll be so much better! No, really!"
In Soviet Rush, today's Tom Sawyer gets high on you.
I just wanted to point out that if we accepted 11Mbps and 54Mbps as the speeds of 802.11b and 802.11a/g then we would have to call regular fast ethernet 200Mbps. 802.11b is 5.5Mbps full-duplex and a and g both are 27Mbps full duplex. It is true that the radio signals are capable of carrying 11 and 54 respectively, but half of this bandwidth is dedicated for each direction, so that the MAXIMUM one-way speed you can achieve with 802.11a/g is 27Mbps. This means that if they're hitting real-life numbers of 24Mbps (I doubt it) of data throughput, then they're doing really well - about 88% of theoretical. That's as good as you can really expect from wired networks, in terms of throughput to bandwidth ratios.
I did not design this game/I did not name the stakes/I just happen to like apples/And I am not afraid of snakes-AniD
Hmm, well their protocol name is clearly hexadecimal: 802.11a = 8*256 + 2 + 1/16 + 1/256 + 10/4096 = 2050.068848
I think you're a bit confused about how the ways are going. At any given time, they're developing several different specs for wireless communication. Some of these, such as the publicized 802.11a, b, and g are hardware-side, meaning they have to deal with the way these things are actually transmitted. Others are more software-side, meaning they have to do with encrypting data and whatnot. Furthermore, all of these (except for a few earlier strange circumstances such as 802.11a) are backwards compatible.
In short, the hardware you buy today WILL be usable in a year or two. It just won't be the fastest, bestest thing on the market. Think of it as Moore's Law translated to wireless communication.
Will they eventually go to 802.12? Or will they keep adding letters (and numbers?) to the suffix? 802.11xczvhyewuohzfhfvoizvuyoiyeoinovhzovcz is sure to be the best protocol evar!
Maybe they should concentrate on approving the 11g standard first.
Just to comment on the "users of 11g" stuff: The implementations you are currently seeing in the shops are based upon more or less early *drafts* of the standard. The fit will really start hitting the shan when people start combining devices from different manufacturers: Incompatibilities range from different modulation schemes (TI) over incompatible MAC protocol elements (dataset identifiers in AP capabilities) to legacy support (some old legacy devices refuse to associate with APs supporting the new modulation modes due to excessive supported-rates-list lengths).
The origin of the slow effective transfer speed by the way is the MAC layer timing. Each information frame is transfered independently and ACK'ed by the receiving station on MAC layer -- the time delays in these frame exchanges take up enough time to reduce the effective transfer rate to about half for the 54Mbps mode. Besides QoS 11e will introduce burst frame acknowledge which should improve the situation considerably and therefore the new modulation schemes may actually make sense.
I dunno if its THAT bad. Personally, I hate upgrading my wireless infrastructure every 8 months it seems...
Why don't those guys forget about trying to make things faster, and instead make things more secure. I have 802.11b but I shut it off when I don't use it because I'm hella scared that my neighbor is going to download kiddie porn off my wireless connection, but the feds will bust me because it looked like it was coming from my ip address!
Let's hunker down, get some real security, and then move to something faster!!!
We need less latency and less packet loss! OK, so the latency isn't that bad with 802.11, but it could be better. Packet loss is the serious problem with 802.11. I've wasted about $2K so far on my home network trying to make it reliable. My DSL router is about 200 feet from my desk, and no matter what I try, I still get nearly 3% packet loss. I've tried four different cards (about $90 * 2 one for each end * 4 # I've tried = $720!), different placements of the antennas, bought parabolic dishes for each end ($300 w/ cables), and spent about 200 hours tweaking. The only thing that really made it better was buying a desk chair with a wooden frame. When I moved my metal framed chair around, the wireless would quit working for a few seconds. It was about 2' behind my dish, but it still screwed-up the reception. With the directional dishes, I've got the noise down to below what the cards can measure (-149dB), but the signal to noise ratio is still only 16 dB. The only obstruction in my line of sight is a single pane of glass in front the dish in my office. I've got my DSL router on my roof with a dish on the other end. I don't know how long it's going to last up there, but I guess I'll buy yet another one when it goes.
It also quits working when I use my cordless phone, but that's understandable since they use the same frequencies. My digital cell phone makes the packet loss worse when it's within a couple of feet of the back of the dish. Unfortunately with the way my office is arranged, I can't rearrange my office to move the chair away from it, so I have to stand while talking on the phone. What I don't understand is why my lawn mower knocks it out completely. My wife can't work on the computer while I'm cutting grass.
It's frustrating the garbage the vendors are putting-out. Obviously with hindsight, I should have just ran a cable. The two parts of my house are on different power transformers, so I was reluctant to do it. When I tried coax Ethernet 13 years ago, the cards wouldn't last a month before frying. If they want us to buy more of the equipment, how about adding reliability rather than features?
If you can find an access-point that will do 802.1x properly with rolling keys, you are in great shape, even with RC4 WEP. The WEP attacks that have been published exploit a vulnerability with poorly-chosen IV values, and if enough encrypted packets are captured, the keys can be inferred. But, if you rekey on a frequent basis, and use a "modern" implementation of WEP which avoids the weak IV values, then you will be fine. Unfortunately, I don't know which SOHO access-point devices currently support an optimum 802.1x implementation. On the client side, Windows XP SP1 does it right, and I believe that Meetinghouse has client implementations for Linux and Mac OSX.
It's just the physics of it.
10Mbps ethernet is 10mbps because that's how many bits per second the channel itself can hold. It actually means nothing at all about the host to host bandwidth. In the case of ethernet, the numbers are very close, so nobody really thinks about it.
11Mbps wireless is the same thing.. it's the bitrate of the radio channel, not the useful data rate of the protocol itslef... and with wireless, there is more overhead. A lot more.
If these numbers are not satisfactory, what number should be reported? It's the only number that is actually standard.. everything else depends on the conditions of the setup, network, and protocols in use.
- Technically we call it "100baseT Full Duplex"
- The 100 refers to 100 bits/second as a maximum channel capacity, not the maximum transfer rate between two hosts. it takes multiple hosts using the channel at the same time to saturate the channel.
- Half of the bandwidth of 802.11b is NOT set for " each direction". The full amount can be used for either direction.. it's half duplex. Further, the 11mbps refers to the radio channel, not any " direction".
- Full channel usage happens with multiple hosts, not with only two. with two hosts.. just like ethernet, but the delays and wait times are larger, adn there is more protocol overhead, due to the lack of collision detection.
That is nice, but by then we will be 10GibE copper
Dumb question... is it possible to install two or more wireless NICs, enable support for splitting tcp/ip traffic amongst them, and get twice the throughput? I thought this was possible using ethernet cable... but I'm not sure how the wireless frequency bandwidth would limit something like this.
What should an average joe like me who wants to have his laptop access files and the 'net onto the main home 'puter should choose ? Is it the same thing that for going on a road trip and hoping to find access points ?
Non-Linux Penguins ?
For virtually all consumer applications, 60% (or less!) of 802.11b's throughput is an ample plenty.
Why? Because the VAST majority of data they schlep goes through their broadband provider on the way to or from the Internet, where they don't get anything even APPROACHING 5 Mb/s. The line out of the house is the bottleneck in my home and in many, many other homes.
This increase in efficiency will be achieved by having slashdot duplicate stories filtered out at the transport layer.
The simple way to get a better throughput on your wireless network is to make it all token ring... after all who is going to have more then five computers in their house/neighbourhood?
and maybe we can get enough bandwidth for wireless VGA! Should only need 600 or 700 Mbps...
802.1x is only half the picture. To run 802.1x you'll need to select an EAP (extensible authentication protocol) method.
Currently, there's 4 common flavors of EAP: EAP-TLS, EAP-TTLS, LEAP, and PEAP.
LEAP is Cisco proprietary and will probably be dumped when Cisco moves to PEAP. Surprisingly, Apple licensed LEAP from Cisco, but only on the client side. You can get LEAP support via the latest update to the Airport client software for MacOS 9 or X. LEAP's weakness is that you can see the username in cleartext over the air, which opens you up to brute force attacks on weak passwords.
EAP-TLS is microsoft's original foray into wireless security. It requires both the radius server and the clients have public key certificates so they can mutually authenticate each other and pass authentication information through a TLS tunnel. Realistically, only corporations with an already implimented Public Key Infrastructure will want to consider EAP-TLS. EAP-TLS is a major administrative nightmare because it requires user certificates, but this also makes it the most secure.
EAP-TTLS was introduced by Funk Software makers of Steel Belted Radius. It only requires that the radius server have a public key certificate. EAP-TTLS builds a TLS (think SSL) tunnel between the radius server and the user to pass authentication information and then inside the tunnel you can use another normally unsafe authentication method such as PAP which sends username and passwords in the clear. In this case though, they're inside an encrypted tunnel.
PEAP was originally a colaboration between microsoft and cisco and does basically the same thing at EAP-TTLS. Microsoft and Cisco's clients no longer interoperate and the whole standard is a mess. Making EAP-TTLS that much more inviting. However because Microsoft control the client and it giving it away for free in XP, PEAP will probably become the standard of the future--the joys of a operating system monopoly.
All of these 802.1x solutions, however, require a radius server to authenticate to. OpenRadius, you think... great if you want to do EAP-TLS. Here's the steps:
1. Install OpenRadius.
2. Install OpenSSL.
3. Setup your server as a Certificate Authority.
4. Issue and sign a certificate for your radius server.
5. Issue and sign certificates for each user.
6. Install an 802.1x supplicant on each client.
7. Install the clients' signed certificates.
8. Configure 802.1x authentication on the access point.
9. Make sure to set up rotating WEP keys, 40 bit WEP really is secure enough if you rotate your keys every half-hour.
802.1x and EAP all sound like wonderful things, but they do not address the needs of the common home user.
If you really want to secure you're home network you should be installing only 802.11a accesspoints and walking the perimeter of your home making sure the wireless signal does not permiate the walls. 5Ghz is a better range for home networking since your cordless phone, wireless keyboard/mouse, and microwave won't cause interference.
I've seen people with new children before, they go from ultra happy to
looking like something out of a zombie film in about a week.
-- Alan Cox about Linus after his 2nd daughter
- this post brought to you by the Automated Last Post Generator...