Mesh offers many routes to network gains

An idea as old as the internet could revolutionise communications

An idea as old as the internet could revolutionise communications

Networking topology is one of IT's least dynamic areas. Major developments included the introduction of the bus topology, the explosion (and subsequent implosion) of Token Ring, and the evolution of the star configuration. All that took place over the past 35 years: it is hardly technology at the speed of thought. But sometimes old ideas find new life and revolutionise an industry. Mesh networking might be one of them.

Mesh networking stands apart from these other network topologies because it does not rely on a single route through the network. Instead, multiple routes are open to it. In a mesh network, nodes are connected together independently of a separate central point, creating a matrix of different links. When a packet is sent over a network, it can be routed in different ways, depending on the algorithm used.

In a mesh network, the number of connections can vary. A fully meshed network would have every node connected directly to every other node. This makes the cost increase exponentially every time a new node is added, says David Tansley, telecoms and technology partner at Deloitte, because you have to add as many extra connections as there are nodes in the network.

There is an alternative, he explains, 'A hybrid where the core of the network is partially meshed. Some connections will have dial-up connections as back-up, and at the centralised command and control level there will be policies and rules that monitor and intervene.'

Mesh networks have two key benefits: resilience, because if one route fails another can be found; and better protection against data overload, because traffic can be routed around congestion. Sound familiar? It should, because the mesh networking concept has existed since at least the late 1960s: the internet is one big mesh.

But companies are considering establishing their own mesh networks. In particular, Nokia has built mesh networking into the IP virtual private network routers that it sells to enterprises for their wide area networks. According to Bob Brace, vice-president of mobile solutions at Nokia, it could make communication between offices more resilient.

Traditionally, VPN connections between a branch office and head office would all be direct, explains Brace. 'After all, what is the point in linking together two branch offices if there was no traffic between them?' he asks. But using the mesh networking technology built into Nokia's router, it becomes advantageous to do so.

The routers use Meta-Hop, a technology from Nokia that enables them to communicate with each other automatically in a self-healing configuration. Normally a branch office with a connection to head office will lose connectivity with head office applications if the VPN route between the two offices fails.

Using Meta-Hop, the connection can be rerouted automatically via another branch office that is part of the mesh, without an interruption to the service. Brace says, 'You simply establish the VPN from that device back to HQ, but then you also establish a VPN link to the other offices if you need it.'

With IP VPNs able to handle multiple connections through different ports, it becomes possible to increase the resilience of your network by maintaining different ISP or ISDN links. And the ability to establish multiple routes back to head office becomes particularly useful in the event of traffic overload, says Brace. Being able to send traffic along a less congested route Ð perhaps via another office which has a clearer link Ð could make data transfer easier during peak periods.

Meta-Hop also includes a self-learning capability, which automatically distributes new IP resource information to all IP VPNs in the network. So, if a new application server is installed at head office, every node in the mesh receives the necessary IP information, making configuration easier.

Mesh networks are usually discussed in terms of Wans, but wireless mesh implementations are gaining appeal as an option in metropolitan area networks or even large campus configurations. One of the advantages of wireless mesh networking is that it is more flexible to configure and easier to connect, because no cabling is needed.

Researchers are already looking into it and pilot schemes are running. Microsoft has been working on mesh network technology for self-organising ad hoc community mesh networks, and MIT has a 50-node rooftop wireless mesh in operation.

Trevor Bicknell, head of management consultancy Arthur D Little's telecoms, information, media and electronics practice in the UK & Ireland, believes that companies and public authorities are only just beginning to explore the possibilities of wireless mesh. 'Companies are thinking more laterally about what objects could be used to provide a facility for mesh networking,' he says. 'Could you use traffic lights?'

If that sounds far fetched, think again. Deloitte's 2005 Technology Predictions report includes the first wireless mesh networks in urban centres. Also, Philadelphia, the fifth largest city in the US, wants to provide free internet access to its 1.5 million people. It will use a Wi-Fi mesh network that will cover its 135 square miles. Traffic lights are likely to provide some of the power for wireless routers.

Stationary objects such as traffic lights may not be the only providers of future wireless mesh capability for corporate users. Many users would be nervous at the thought of their laptop, PDA or smartphone acting as a router in a larger network fabric. But Tansley sees mileage in the idea, especially if users control data exchange criteria.

'You may have a group of people entering a meeting room where they all have Bluetooth-enabled PDAs,' he says. 'If they connect, and they all decide that they want to participate, then for the duration that those devices are in close proximity, they will either exchange information or not.'

Although it raises security considerations, this idea is appealing, because resilience comes built-in. If a device fails or someone walks out of range, the data simply reroutes.

Now that the internet has proven the concept so well, mesh networking has potential at all levels: from cabled corporate Wans with branch offices, through to wireless metropolitan area networks and campuses. Mesh networking is still in the early stages of commercial exploration, but as things start to unfold, corporate IT departments are likely to untangle many more of its possibilities.

www.research.microsoft.com/mesh

www.pdos.lcs.mit.edu/grid

 The world of wireless mesh networks

Wireless mesh networks are an alternative to wired mesh networks, and also 802.11-based wireless local area networks.

Wireless mesh networks make use of redundant connections between wireless nodes to create a self-configuring and self-healing network. Each node has intelligent software in it to direct data packets from one device to another, so they can travel efficiently to their destination.

Just as the US Department of Defense played a key role in the development of the internet, mesh networks were originally developed for military applications. In both cases, the rationale was to develop a network that was highly robust.

Pros and cons of wireless mesh

Pros

  • Mesh networks do not require a line of sight.
  • The load on any given access point is not as great as it would be with a wired network. Because there are many ways to route a packet, it is not necessary for all packets to go through a single access point, just because a number of users are near a specific access point.
  • Roaming is seamless because routing information is always available to authenticate a user.
  • Cabling requirements are minimal, and only one device on a segment generally needs to be tied to a wired network.
  • Distance limitations do not apply to mesh networks, because each device acts as a router and as a repeater.

Cons

  • According to Forrester, the cost of creating a metropolitan wireless mesh network is far more expensive than a traditional 2.5G network
  • There is no standard for mesh, and there is not likely to be one for another three years.
  • Scalability is an issue. There is currently no installation large enough to determine whether bandwidth could be overwhelmed by too many routing information updates.
  • More users from different sectors are needed to prove the technology. Public safety and local authorities have been the early adopters.
  • There are too many start-up suppliers, and Forrester believes IT managers should be wary of supplier consolidation in the future.
This was last published in April 2005

Read more on IT operations management and IT support

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.

-ADS BY GOOGLE

SearchCIO

SearchSecurity

SearchNetworking

SearchDataCenter

SearchDataManagement

Close