Wireless point-to-point or point-to-multipoint networks supplement campuswide Ethernets, bridging networks and avoiding a backhoe problem, but otherwise they act much like a line-of-sight wire or a high-tech tin-can-and-string network.
Although point-based wireless networks that use unlicensed radio bands in the 2.4GHz and 5GHz range are often inexpensive and uncomplicated to install, they quickly max out their potential when line-of-sight obstacles stand in the way or the number of points exceeds a relatively low limit.
The successor and perhaps ongoing complement to point-to-point is a mesh network. In many ways, a mesh network resembles an idealized version of top-level Internet backbone routing in which physical location is less important than capacity and network topology. Only one or a few nodes need to be connected to a wired backbone or some sort of network uplink, and each packet bound for the same destination might proceed along a different sequence of nodes.
In point-to-point networks, each link requires a dedicated connection; in point-to-multipoint set-ups, the multipoint transceiver winds up being the bandwidth bottleneck and the point of worst failure. A mesh can distribute bandwidth and risk, while reducing the cost of deployment and expansion.
Depending on the implementation, each mesh node has a degree of autonomy or algorithmic intelligence that guides how it optimizes paths from instant to instant, whether in mobile networks or conventional networks in which the load shifts dynamically.
In the ideal mesh network, nodes provide automatic failover, locking out dead or nonresponsive peers, and routing is revised without any manual reconfiguration when new nodes are added.
Mesh networks divide almost neatly into two types: fixed wireless installations that connect multiple locations and mobile, peer-to-peer, ad-hoc networks that have variable availability and a potentially everchanging set of members. One also could deploy mobile or temporary networks that have the same properties as a fixed network.
Fixed mesh networks are generally built with the expectation that many nodes have no direct backhaul, network, or Internet access. In fact, if each location had some kind of enterprise or Internet access, distributing service by wireless would be almost unnecessary.
In a fixed installation, locations for nodes are chosen with an eye for providing the right overall level of bandwidth with the fewest points. Fixed mesh networks also can effectively offer non-line-of-sight service by ringing an obstacle -- a tall building, a hill, a cluster of trees, an area of known interference -- with enough nodes to bypass it. These fixed networks are typically directional enough over each link to avoid major security risks.
In contrast, peer-to-peer mobile mesh networks -- which are a long way from actual deployment -- rely on individual devices connecting to each other through devices within radio range. Scalability can be an issue because each device has to manage known optimal paths, which can change from millisecond to millisecond. When an uplink of some kind is added via cell, satellite, or wire, the network becomes dynamically aware and can handle queued interactions.
Security is of paramount concern in mobile meshes, although the technology's use in military battlefield projects provides some security through obscurity. Because of the ad hoc nature of mobile mesh, most of its use would be in the wild, outside of enterprise control.
Several companies already offer commercial products that provide fixed mesh networks for residential/business broadband service in both the 2.4GHz and 5GHz unlicensed bands. The advantages of deploying these solutions benefit both wireless ISPs and their customers by extending service areas around line-of-sight limitations and reducing costs. Each new customer becomes a beachhead for further network expansion.
Companies targeting service providers with tower-located equipment and CPE include Cowave, LocustWorld, Radiant Networks, and SkyPilot. Nokia Rooftop, an early wISP mesh product, has been discontinued, according to a Nokia representative.
LocustWorld uniquely offers its firmware under an open-source license and sells many of its products at cost, reserving fees for commercial projects that deploy its products or software.
In the enterprise world, FHP Wireless seems to stand alone in two respects: It is the only company offering a mesh system and combining routing with standard Wi-Fi in each node.
In pure mesh systems, each node routes traffic across the mesh and bridges it to backhaul or a local network. In FHP’s system, nodes route traffic but distribute it through an integral Wi-Fi station. If deployed densely, these systems create a “hot zone” with reduced wiring cost and great flexibility in increasing density or changing coverage.
One FHP product type, currently called SmartPoint, handles mesh routing and Wi-Fi distribution, while another, currently called RoutePoint, bridges backhaul into the system, allowing bandwidth to be added to a mesh in any location. (The company is in the process of rebranding both products.) FHP’s approach relies on existing Wi-Fi client infrastructure.
FHP may soon face competition from the biggest fish in the sea: Intel put mesh on its radar at its developer conference in mid-February at which it demonstrated mesh technique and discussed its interest in LAN and WAN deployment. Intel also floated FHP and other companies’ boats simply by stressing the credibility of mesh as an alternative.
MeshNetworks is apparently the only company to have deployed a mobile mesh product currently on the market. The company's peer-to-peer system requires special client and server hardware that works like Wi-Fi with extensions
MeshNetworks' client hardware, the PC400 PC Card, talks both plain 802.11b and mesh interchangeably. MeshNetworks says that instead of ratcheting down to a slower rate to speak to a distant access point, its client adapter can talk at a higher speed to another mesh client adapter. MeshNetworks also offers the WR400, a repeater for fixed enterprise deployment, and the AP400, an access point that bridges both mesh clients and standard 802.11b clients to a network.
Mesh networking's time has arrived, but real deployment is still to come. Although products are increasingly real, not just academic, few companies have experience with practical performance. But with Intel signaling its intent to jump into the market, mesh should become another practical tool in the campus network kit.
Many enterprises could benefit from expanding their networks through a mesh by reducing single-link dependency and expanding coverage area. In mobile deployments, mesh networks could bottleneck in a too-dense environment as each low-powered handheld maintains traveling salesmanlike routing tables.