Communities experimenting with mesh networks(technical.ly)
technical.ly
Communities experimenting with mesh networks
http://technical.ly/2015/04/06/12-communities-experimenting-mesh-networks/
3 comments
Does anyone have any opinions on how this could grow to replace traditional ISPs? What's the difference between a mesh network and just putting a big WiMax antenna up in my neighborhood and giving everyone free internet access?
The idea of a meshnet is that, as the name implies, it's a mesh. It usually involves intelligent rerouting over other peers if a previous route is no longer available.
If you were to put up a big WiMax antenna, you are the single point of failure - you could spy on the traffic, you could banish specific clients, and if you were to disappear, everybody in your neighbourhood would be without internet.
A mesh network allows none of these (in theory), because there is not one single point of control.
If you were to put up a big WiMax antenna, you are the single point of failure - you could spy on the traffic, you could banish specific clients, and if you were to disappear, everybody in your neighbourhood would be without internet.
A mesh network allows none of these (in theory), because there is not one single point of control.
Mesh is a topology, not a belief system. A mesh network can still be administered by one entity who can monitor traffic, manage configuration, and restrict access to authorized users, but they would do so using a distributed system rather than the control panel of a single piece of equipment.
It's different from what you describe in that a typical WiMax antenna setup mirrors the star topology typical with network switches. Clients are connected to a piece of infrastructure, which is in turn connected to a larger piece of infrastructure, and so on up the tree.
In a mesh network, clients are connected to each other without using infrastructure as an intermediary. Every node relays traffic for its neighbors, and you reach distant points by hopping through many of your neighbors (or you don't, if they don't happen to currently be arranged in a way that facilitates that).
A good mesh network system has intelligent routing algorithms that are some combination of fault-tolerant, efficient at finding the shortest path, equitable in distributing load, and designed well in terms of incentive structure (so that people do not freeload by acting as clients but not relays).
If you are using a mesh network to reach the internet, you are still dependent on the nodes that route between the mesh network and an actual ISP, which presumably pay for the privilege and want to recoup costs somehow. However, it can be a decent way for a community to roll its own last-mile infrastructure, provided it can purchase wholesale transit somewhere within the mesh, and drive enough participation that there is sufficient capacity between you and the backhaul most of the time.
The more interesting uses are for meshes that are not (mostly) about Internet access. There is experimental work on protocols for messaging and durable document storage/publication (think Wikipedia but without web servers) that work well on meshes, providing pretty good internet-like capabilities to areas where access to the actual Internet is impractical. (Some of this is embedded in OLPC/XO).
It's different from what you describe in that a typical WiMax antenna setup mirrors the star topology typical with network switches. Clients are connected to a piece of infrastructure, which is in turn connected to a larger piece of infrastructure, and so on up the tree.
In a mesh network, clients are connected to each other without using infrastructure as an intermediary. Every node relays traffic for its neighbors, and you reach distant points by hopping through many of your neighbors (or you don't, if they don't happen to currently be arranged in a way that facilitates that).
A good mesh network system has intelligent routing algorithms that are some combination of fault-tolerant, efficient at finding the shortest path, equitable in distributing load, and designed well in terms of incentive structure (so that people do not freeload by acting as clients but not relays).
If you are using a mesh network to reach the internet, you are still dependent on the nodes that route between the mesh network and an actual ISP, which presumably pay for the privilege and want to recoup costs somehow. However, it can be a decent way for a community to roll its own last-mile infrastructure, provided it can purchase wholesale transit somewhere within the mesh, and drive enough participation that there is sufficient capacity between you and the backhaul most of the time.
The more interesting uses are for meshes that are not (mostly) about Internet access. There is experimental work on protocols for messaging and durable document storage/publication (think Wikipedia but without web servers) that work well on meshes, providing pretty good internet-like capabilities to areas where access to the actual Internet is impractical. (Some of this is embedded in OLPC/XO).
Mesh networks are suitable when all the computers you need to talk to are in a geographically contiguous area. As soon as you start needing to cross oceans or mountain ranges, you're likely to run off the edge of a mesh.
So I don't anticipate meshes replacing traditional ISPs any time soon.
So I don't anticipate meshes replacing traditional ISPs any time soon.
Pure mesh networks will not take over ISP's. But what about a hybrid solution? Fat fiber pipe feeds wireless mesh. Run fiber from nearest datacenter to middle of town, build mesh network out from there.
Crowdfunding SEC rules + neutrality ISP rules could very well open up an opportunity for crowdfunding municipal ISP's. Communities run their own mesh and rent fiber from an upstream transit probider.
Crowdfunding SEC rules + neutrality ISP rules could very well open up an opportunity for crowdfunding municipal ISP's. Communities run their own mesh and rent fiber from an upstream transit probider.
A mesh solution is workable, and possibly cheaper than a traditional deployment model. However, I can see significant drawbacks. It's likely to be less reliable, slower, and not well suited to anything requiring low latency. It's also likely to be more prone to congestion problems and will probably be more difficult to manage.
So it's possible, but I don't think it's a good model for how most Americans are likely to be using their internet connections. Most of their use would likely be the mesh as transit to the uplink and then off to Facebook or Amazon or Netflix.
I think the lesson is that meshes and traditional systems have different strengths and weaknesses. One isn't always a good substitute for the other.
So it's possible, but I don't think it's a good model for how most Americans are likely to be using their internet connections. Most of their use would likely be the mesh as transit to the uplink and then off to Facebook or Amazon or Netflix.
I think the lesson is that meshes and traditional systems have different strengths and weaknesses. One isn't always a good substitute for the other.
How is this different than the current situation, other than one is wires, and one is wireless? ISP's still need to aggregate household cables into larger fiber before hitting the datacenter.
A wireless ISP is designed around high-bandwidth uplinks to aggregate from each tower/AP to central points. A mesh doesn't have this topographical feature. (Other considerations: limited frequency bands)
The difference, I think, is that the traditional layout works better for the common use case than a mesh does. A mesh with an uplink would be preferable in a scenario where a traditional system is too difficult or expensive to deploy, or where traffic is going to be generally internal.
Oh, and where you have some kind of trust mechanism in place. A mesh with malicious nodes is potential havoc. Central authorities give you a chain of trust that mesh systems don't have.
And no, doing away with that central trust for political reasons is generally not as good an idea as it sounds like. Managing trust without any central points is much more difficult and much easier for users to screw up.
The difference, I think, is that the traditional layout works better for the common use case than a mesh does. A mesh with an uplink would be preferable in a scenario where a traditional system is too difficult or expensive to deploy, or where traffic is going to be generally internal.
Oh, and where you have some kind of trust mechanism in place. A mesh with malicious nodes is potential havoc. Central authorities give you a chain of trust that mesh systems don't have.
And no, doing away with that central trust for political reasons is generally not as good an idea as it sounds like. Managing trust without any central points is much more difficult and much easier for users to screw up.
Oh, I see. I was using "wireless" and "mesh" interchangeably, but you are correct in distinguishing between them. There are routing differences between "wireless," where one company owns all the wireless routers, and "mesh," where ownership is decentralized and therefore so is routing.
There could be some middleground we are not considering. A "federated" solution could be a good answer to this. Multiple companies/individuals provide the intermediate nodes and are responsible for their maintenance (and hopefully incentivized a la Torcoin). The trust issue is hard to solve, but variations of onion routing protocols exist that ensure only the exit node sees unencrypted traffic. At that point you are still trusting one entity, but trust is no worse than status quo of ISP's.
Also, your post seems to imply wireless ISP (buy uplink, route over wireless) is already a thing. Is this in the USA? I would love some examples.
There could be some middleground we are not considering. A "federated" solution could be a good answer to this. Multiple companies/individuals provide the intermediate nodes and are responsible for their maintenance (and hopefully incentivized a la Torcoin). The trust issue is hard to solve, but variations of onion routing protocols exist that ensure only the exit node sees unencrypted traffic. At that point you are still trusting one entity, but trust is no worse than status quo of ISP's.
Also, your post seems to imply wireless ISP (buy uplink, route over wireless) is already a thing. Is this in the USA? I would love some examples.
The "middle ground" you describe is how internet is already provided in the US. Multiple companies provide intermediate nodes, connectivity, and bandwidth. Traffic is routed and re-routed according to network management protocols. They are incentivized to do this via USD. Trust is handled via centralized authorities and PKI.
Wireless ISPs do exist in the US. monkeybrains and Clear both come to mind.
Wireless ISPs do exist in the US. monkeybrains and Clear both come to mind.
What I think might contribute to success of mesh networks as ISP alternatives is incorporation of micropayments for bandwidth/storage into the system.
There should be an incentive for individuals to participate. A thread of catastrophic communication system failure (as in case of a natural disaster) is not enough to stimulate the deployment. But the promise of income is.
Would you be inclined to install a mesh node if it gave you a small income stream?
There should be an incentive for individuals to participate. A thread of catastrophic communication system failure (as in case of a natural disaster) is not enough to stimulate the deployment. But the promise of income is.
Would you be inclined to install a mesh node if it gave you a small income stream?
How would the income system work? (..cryptocoins?)
What goods/services can people buy with your "income stream" system?
What goods/services can people buy with your "income stream" system?
Storage, bandwith, voice.
The problem, I think, is distributed accounting - in the presence of untrusted and malicious nodes. Micropayment transactions need to work faster than existing systems like BTC, to be able to support fast call set up, for example.
The problem, I think, is distributed accounting - in the presence of untrusted and malicious nodes. Micropayment transactions need to work faster than existing systems like BTC, to be able to support fast call set up, for example.
I love this stuff.
I am taking a good Coursera course "Fog Networks and the Internet of Things" that covers edge networks. I am also planning on including a section "IoT and Edge Fog Networks" in my new book I am writing.
There is a lot that can be done at the edge (collecting and processing local data, sharing computational resources, etc.) without needing LTE/core or the cloud.
I am taking a good Coursera course "Fog Networks and the Internet of Things" that covers edge networks. I am also planning on including a section "IoT and Edge Fog Networks" in my new book I am writing.
There is a lot that can be done at the edge (collecting and processing local data, sharing computational resources, etc.) without needing LTE/core or the cloud.