Introducing 802.11s - Wireless Mesh Networking

ikewillis writes "Intel has introduced a new wireless networking standard called 802.11s. This standard utilizes a mesh topology, allowing for fully self-configuring networks where each node can relay messages on behalf of others, thus increasing the range and available bandwidth with the number of nodes active within the system, versus the point-to-point structure of existing WiFi networks. This will radically transform WiFi hotspots, allowing the geographical area and available bandwidth on the network to scale with the number of participants."

This is great but...

WiMax and other technologies like it will still be much more important because, do we really want a grid of short range networks that will ultimately cause divisions between different parts of the networks if one node goes down or would we prefer enourmously large networks that overlap each other (the different nodes) once or twice or thrice?

Wimax is LICENSED, Wifi is NOT licensed

once you get licenses in the picture, you disempower the smaller entities and empower the larger entities. And I think that most Americans are starting to see that whenever larger entities gain power over small entities and citizens, then things start to go sour...

WiMAX runs over unlicensed freqs too

Just an FYI, WiMAX runs across both licensed and unlicensed bands.

Re:Wimax is LICENSED, Wifi is NOT licensed

Frankly, I have a certain suspicion that most Americans (perhaps even more than other nationalities) are too busy watching wrestling, praising Jesus, declaring war on abstract nouns, etc., to pay attention to whether the leader of their country is capable of rational discourse, let alone whether a particular wireless protocol is empowering large companies or not.

(Yes, I'm going for an even split between funny and troll).

Re:This is great but...

Actually, available bandwidth can increase with users in some situations. It depends on how many bands are available, how many landline connections, topology, etc. Lots of factors to consider.

As a trivial example, consider two networks, one with mesh one without

A net1 B mesh C net2 D

Bandwidth from A - D is the minimum(net1, mesh, net2).

versus:

A net1 B nothing C net2 D

bandwidth from A - D is 0.

As a slightly more complex example: /-mesh1-B-\
A--D
mesh1 \-mesh2-C-/

Is the bandwidth from A-D more or less with or without C?

Making hardware do what people expect it to do

Have you ever installed one wireless access point, and wished you could install a second, within wireless range of the first, without running a second cable? Most access points can't do that, even though most people expect them to be able to before being told otherwise. Mesh networking would enable this sort of networking, and much more.

The performance will always be less than an "every AP has its own landline" topology, but networks will be much easier to build (and perhaps simpler to maintain).

Re:Making hardware do what people expect it to do

Have you ever installed one wireless access point, and wished you could install a second, within wireless range of the first, without running a second cable? Most access points can't do that, even though most people expect them to be able to before being told otherwise. Mesh networking would enable this sort of networking, and much more.

Are you talking about a repeater? I believe most of the cheap linksys APs can be set up to be repeaters instead.

s?

where do they get all these letters from? There seems to be 802.11b, 802.11a, 802.11g and now 802.11s, and I have no idea why the letters are what they are. Anyone care to explain?

Re:s?

No. We will not explain.

However, we will chide you for not including 802.11n on your list!

Goofy letters

http://computer.howstuffworks.com/wireless-network 2.htm [howstuffworks.com]

Check out the whole article to find out more about the various 802.11x standards (excluding the new 's' one).

Re:s?

What do the letters mean?

"Task groups within the 802.11 WG enhance portions of the 802.11 standard. A particular letter corresponding to each standard/revision, such as 802.11a, 802.11b, and so on, represents the different task groups. For example, Task Group B (i.e., 802.11b) was responsible for upgrading the initial 802.11 standard to include higher data rate operation using DSSS in the 2.4GHz band."

From 802.11 Alphabet Soup [wi-fiplanet.com].

Re:s?

to tell the truth, 802.11a, 802.11b, 802.11c, 802.11d, ... exists, but some are less used (and known) than others.

More info (with explanations) here [wi-fiplanet.com]

Re:s?

802.11a = apathy. No one really paid much attention
802.11b = bad. It works but there's better to come
802.11g = good. Now it's worth using
802.11s = shit. That's what users on the fringe of the network will be screaming when the "link" node between the access point and them finishes their lunch and leaves, cutting them off too

Re:s?

802.11a = apathy. No one really paid much attention

You're right there, and this makes 802.11a a great thing to have. I'm running my home network on 802.11a, and here are the benefits I reap versus 802.11b/g:

1. When the hardware was available but on the way out, it was -very- cheap to pick up;

2. The range is much more limited than b/g, but big enough to cover my house and backyard, so I have less worry about "sharing" my connection with my neighbors than with b/g;

3. The 802.11a range is underutilized (my neighbors don't have 802.11a, and yours probably don't, either) and doesn't shut down by interference when you use the microwave;

4. Someone wardriving or just playing around with wireless sniffing tools from their bedroom are much less likely to be using 802.11a; in fact, until recently airsnort and related tools didn't even have 802.11a compatibility, and getting 802.11a working with Linux is a PITA compared to 802.11b/g.

So in a way, using 802.11a improves your odds of a secure and non-shared connection in the same way that using Opera improves your odds of picking up a javascript exploit from a web site. That's not security in and of itself, but coupled with VPN and the reduced range, it's very nice indeed.

Can do with existing protocols

Well, mesh networking does not necessarily need a new 802.11x spec. This article [tombridge.com] on Tom Bridges blog [tombridge.com] is republished from the first issue of Make [makezine.com] outlines how to create mesh networks using an Airport Express.

Too bad cities won't be able to do it.

The way things are going, cities won't be able to provide this for their citizens. No one needs a network this big for personal usage; if municipal wi-fi is banned, it will be for naught.

Nifty . . Highway net!

I think it sure would be nifty to see this type of AP installed in cars and have uplink points along major highways . . . It'd be a fluid network that would improve with traffic . . . Then again, maybe encouraging heavier traffic is a bad thing . . . it'd still be cool.

A clarification and question

Intel has not introduced the 802.11s standard; Intel has made a proposal to the IEEE, which they will take into consideration while designing the 802.11s standard.

The article makes 802.11s sound like a general mesh standard, which would be really nice. However, what I read on the IEEE Web site recently made it sound like merely a self-configuring version of WDS (so that only access points participate in the mesh). Can anyone provide details on the features of Intel's proposal?

Re:A clarification and question

However, what I read on the IEEE Web site recently made it sound like merely a self-configuring version of WDS (so that only access points participate in the mesh).

Yes, that's basically the idea behind the 802.11s Task Group-- but the phrase "self-configuring version of WDS" really doesn't quite go far enough in describing the concept. It's sort of like describing the Internet protocol as a "self-configuring version of frame-relay". Probably not helpful.

Wireless mesh networks are multi-hop in a way fundamentally more complicated than the simple access point and a bunch of associated stations. They'll have to run a routing protocol and forward from mesh node to mesh node in an efficient and secure way. They'll have to be robust in the face of individual node failure. They'll have to support stations roaming securely between nodes in the same mesh network. It's a whole lot more then just self-configuring WDS.

Folks shouldn't get too excited about this standard. There are a lot of obstacles to making large multi-hop 802.11 networks as efficient as similarly wired topologies. The 802.11s task group isn't chartered with fixing the problems in the MAC layer that keep multi-hop networks from scaling up to very large meshes.

What are the problems? The big one is that they have a profoundly negative effect on TCP fairness. Next up is that multicast is just horrible. Even on regular 802.11 infrastructure networks, it's just horrible. On mesh networks, don't be surprised if it's even worse.

FreeMeshWeb?

Set up enough of these, and you could do your own neighborhood network...

Could this jump-start the "freeweb" movement, particularly since the telcos are lobbying and pushing to kill the muni wireless attempts?

Let's get the entrepreneurs and the networking hippies on the same "frequency."

Sounds great but unreliable?

What happens when a node goes down between several other nodes and the other nodes are now out of range of each other? The network will split and the result will be two seperate networks that are unable to reach each other until the connecting node is up again. Will users be constantly facing problems similar to IRC netsplits? Not to mention that all equipment would need to be replaced to take advantage of this new standard. I'd be more interested in longer range, or more robust signals that can penetrate more obstacles.

WiFi lower level protocol vs. IP

There are already many research projects ongoing which try to find good routing algorithms and network topologies for IP based mesh networks.

Most of these projects try to build their mesh networks on the IP level, i.e. hardware and, IMHO even more important, medium independent.

This standard seems to work below the IP level, i.e. invisible for normal routing hardware and only usable with those "s" devices.

I wonder if this is really a good idea. Making such a standard prevents altering and improving the routing algorithms (because in the best case, they reside on some FPGA) or using mesh network topologies with, lets say, a mixed WiFi, free space optical (think house to house laser pointers :) and ethernet network. You'd need upgrades for a new routing algorithm and progress in this area will be much slower.
OTOH, maybe the network will be more stable, but one has to prove that.

Whither the Internet?

Couldn't this theoretically replace the internet altogether? Once the densities of these "s" hotspots is high enough, wouldn't it be theoretically possible to retrieve a page, send an email, etc. without ever having to transmit the message over the internet "proper"?

This is premature

As someone who does research in this area, I think this announcement is a little premature. There are several fundamental problems that have yet to be solved with this sort of wireless network topology, and I don't see any indication that the 802.11s task force has solved them.

For example, no one has given a MAC protocol that solves the hidden/exposed sender/receiver problems simultaneously. Without such a MAC protocol, it is impossible to resolve the contention fairly. 802.11 DCF solves hidden and exposed sender, but not receiver.

Also, Gupta and Kumar [bell-labs.com] showed that the per-node bandwidth in a wireless mesh with random node placement is O(1/sqrt(n)). This is especially bad news for the sort of nationwide wireless meshes people have been talking about here.

Finally, TCP is especially problematic [ieee-infocom.org] over multiple wireless hops. It causes self-interference which creates massive packet loss due to contention. TCP is built on the assumption that all packet loss is from congestion, but this assumption is not met by wireless contention losses.

In my own simulations, TCP's overaggression causes routing packet losses, creating spurious route breakage and even more TCP timeouts.

Re:Lack of security?

It's worse than that. Most wireless mesh technologies allow anyone to advertise themselves as the quickest route to somewhere, provided they're closer to the source than the real destination. They could then either a)modify the packets (if the nodes aren't using end-to-end encryption) or b)drop them.

This is a good question and, last I checked, an open research topic. One workaround is to only accept route advertisements from a trusted set of routers.