Mae'r cynnwys hwn ar gael yn Saesneg yn unig.

MESH Networks

21 Hydref 2005

How do they help?

Mesh networks have the potential to bring several advantages to wireless communications services, namely:

  • They can allow the formation of a new type of network where users exchange information without the need for network infrastructure. As well as allowing a different commercial model it is often claimed these are more spectrum efficient.
  • They can extend coverage of cellular and other networks by allowing terminals on the edge of the coverage zone to relay signals to those who do not have coverage.

An often quoted vision of mobile communications describes the future as the integration of all mobile and wireless nodes (e.g. cellular, WLAN, PAN etc) with an IP core. One potential application of mesh technology would be to provide another route, alongside WLAN and 3G etc, into such a core network.

Vision of an integrated future

Figure 1 mobile integrated vision using IP core

Mesh Attractions

Perhaps the largest attraction of meshes is that they can be entirely unplanned. This is useful to the military and to disaster recovery teams who desire this ad-hoc networking capability for fast deployment and flexibility in situations with little fixed infrastructure. It is less clear what this benefit brings to the roll-out of a mass market mesh network, although for a service provider or regulator, the lure of a network which promises no planning phase must be high and thus merit investigation.

Another strongly attractive feature is in coverage, where they can offer complimentary performance to that of cellular systems. Meshes have the ability to provide coverage in cluttered environments such as the urban environment. A chain of mesh nodes can ‘hop’ around corners in an urban environment in a way the cellular P2MP systems cannot.

Ofcom is researching mesh networks to:

  • identify the theoretical determinants and metrics of spectral efficiency for both high frequency (line of sight) and low frequency (non-line of sight) mesh systems
  • investigate the capacity constraints of mesh networks and examine the hypothesis that for a mobile mesh the more consumers use a service, the more capacity the network has
  • investigate whether mesh systems have any regulatory impact, e.g. would the wider use of mesh systems imply that there should be more licence exempt spectrum?
  • examine the key problems in the delivery of fixed and mobile mesh systems, understand what is required to resolve these and what the timescales for widespread adoption of mesh might be.

Ofcom is investigating mesh networks to understand the true practical benefits that the technology might bring.

The work we have commissioned in this area is addressing questions such as:

  • Are meshes more spectrally efficient than alternatives?
  • Can meshes enable the use of higher frequency bands, and/or support services-types that alternatives cannot?
  • Are meshes practical, and what are the enabling technologies?

Our work in this area will cover both fixed and mobile mesh applications. The work presented here concentrates on mobile mesh applications, though some of what is said will apply to the fixed case also.

The work has shown that in discussing the performance of mesh systems care must be taken since the type of mesh used and its application can lead to very different conclusions regarding the performance that may be expected.

Capacity and Scalability

Do customers self-generate capacity in mesh?

There are huge attractions to having ‘self generation of capacity’ in a radio network. Notably, that the network is self-sustaining and that it could avoid the so-called ‘tragedy of the commons’. Such a tragedy relates to the days when common land was used for the grazing of livestock with free access for all. The danger is that free access to a finite resource can result in that resource being fully consumed or compromised further such that it loses its usefulness to all. What then, if each user were somehow to add grazing capacity as they joined the common?

The hypothesis that in a mesh network the subscriber base self-generates capacity is crucial for understanding the likely applications and performance of mesh systems. To establish whether this hypothesis is valid in practical applications, four published approaches supporting this standpoint have been reviewed. Each presents a coherent argument based on its stated assumptions, however those assumptions do not translate well to practical applications. The assumptions were:

  • unbounded latency for network traffic
  • unbounded requirements for spectrum
  • confinement of nodes into localised groups in a large mesh

These assumptions place a significant limit on the applicability of the self generating capacity. The work has concluded that for a pure mesh, subscribers cannot self-generate capacity at a rate sufficient to maintain a target level of per-user throughput regardless of network size and population. The only viable ways scalability can be achieved are by providing additional capacity either in the form of a secondary backbone (fixed) mesh network – so forming a “Hybrid Mesh”, or an access network – so forming an “Access Mesh” as shown in Figure 2. In these two configurations scalability is possible and has similar characteristics to that of a cellular network.

Relay nodes provide stucture for a mesh of mobile nodes

Figure 2 Hybrid Network: Intra-Mesh traffic with Infra-structure support

The conclusion from the work undertaken is that meshes have no especially good properties with respect to scaling. In particular as node density and geographic size increase, the traffic rate available to any particular user decreases. The implication of this is that mesh networks should not be chosen over cellular networks on the basis of capacity alone.

This work shows that this lack of scalability can only be overcome by either adding additional capacity in the form of a hierarchical network or containing the end-to-end traffic flows to localised regions within the network

All current theory and measurement of ideal, novel and practical meshes conclude that ad hoc mesh networks comprising only peer-to-peer communication links do not scale well with increasing node population unless there are specific limitations on the density of nodes; the propagation environment and the traffic models.

Additionally there exist practical MAC and routing challenges which further push for meshes which have a low hop count – and hence localised traffic flows.

Underlying causes of limited capacity

Transmissions from nodes in a mesh extend beyond the wanted range to a wider ‘interference zone’, as shown below. Other nodes wanting to communicate within this interference zone must use other elements of time/bandwidth resource. Given that this is a finite resource this can lead to bottlenecks in communications across these interference zones, particularly as node density rises.

Signal passes through wanted signal boundaries and interference boundaries (a)

Figure 3 Mesh node interference (each colour represents a different frequency channel)

Clearly it is advantageous to keep this area, a, as small as possible. This confirms the conclusion of other researchers that short hop lengths and high propagation attenuation factor are conducive to high throughput capacity of the network.

Project status

Our work in this area addresses the role that mesh networking will play in support of our vision of future wireless devices providing high bandwidth connections at home, in the office and outdoors.

The work concludes that meshes work best for the scenario of connections to extra-mesh services such as the telephone network and the Internet. This type of mesh will support applications such as extending hotspots to wider areas, provision of broadband networks and internet to rural communities, or provision of wireless networking at lecture halls or conventions.

Such mesh networks therefore will require infrastructure for deployment in the form of access points to connect to the external network. The work concludes that mesh networks can scale and provide a sustained level of service as new users join as long as the density of such planned access points is kept sufficiently high. This represents a form of ad hoc network in that users may join in an ad hoc manner, but the infrastructure itself must be planned and scaled, very much like cellular networks. Such meshes will not be as quick to deploy as pure intra-meshes, however will still be quicker to install than any new wired or cellular system, so will still have clear benefits as an alternative in some applications. An example application might be deployment to cover a new industrial park or a temporary conference event. Thus we view meshes as likely to play a role in our vision of increasingly mobile communications, supporting the ability for mobile devices to increasingly connect to broadband networks at any location.

For a pure mesh network where there is no infrastructure to provide connections to external networks such as the internet, the benefits of rapid set-up and tear down are accrued. It is in this area that meshes were originally used in defence applications and are likely to find further application in emergency service operations where planned infrastructure is unavailable. However, for this type of pure mesh subscribers cannot self-generate capacity at a rate sufficient to maintain a target level of per-user throughput regardless of network size and population. Thus this type of mesh is unlikely to find widespread commercial application.

The work has shown that meshes are not an improvement in spectrum efficiency in practical cases in comparison for example to cellular networks. Improvements in spectral efficiency of a mesh network can be made through the use of directional antennas, however this is likely only to be available to fixed mesh applications. In the mobile case small handset sizes preclude the benefits of spectral efficiency.

Improved utilisation of the spectrum is possible however since many of the applications for mesh networking can be efficiently deployed at the higher frequencies outside of the congested high demand spectrum.