Wireless local area networks or LANs have been available since the late 1980's, but the market remains immature due to a dearth of standards and predominance of incompatible proprietary solutions.
By definition, LANs are local in terms of networking technology and thus involve none of the complexities of routing, internetwork address resolution, name-to-address translation, segmentation and reassembly of data packets, etc. Also, since traditional LANs have been bound by the limitations of physical media, such as the 500 meter maximum coaxial cable length for a traditional 10-BASE5 Ethernet, user mobility has not traditionally been an important consideration.
The traditional motivation for wireless LANs has been the desire for flexibility of user location within a building or campus. With a wireless LAN, a user's workstation is no longer physically constrained to local network taps. This is particularly effective in environments with frequent user moves, additions and changes (such as point-of-sale terminals in a retail situation) or in situations where installing cabling is either undesirable (such as historic buildings) or unsafe due to construction issues (such as asbestos ceilings). Manufacturing environments, with robot-driven machinery have also found increasing use for wireless communication networks.
Another growing motivation for wireless LANs is the demand for ad hoc user networks, such as sharing of data by attendees at meetings and conferences. Previously this need was met via the proverbial "Sneakernet," which, despite being a wireless LAN technology, will not be discussed further. One could consider ad hoc networks to be a special case of mobility, but aspects of mobility such as security (authentication) and external accessibility have not been an issue, although they probably should be, considering the rapid deployment of viruses which is possible in these settings.
A third motivation for wireless LANs is the desire to remain "in touch" while moving through a local area such as a building or campus. Roaming software extending the effective wireless LAN coverage area has recently become available allowing mobile users to maintain contact while wandering among overlapping cells. Wireless access points monitor the signal strength of moving hosts and coordinate handoffs from one point to another. However, the rate at which a user is allowed to move is typically limited to pedestrian speed.
Three primary physical technologies have been used in wireless LANs: infrared, narrowband RF and spread-spectrum RF. Their characteristics are summarized in Table 9.1. [WONG95] provides a fine summary of these physical technologies. Adoption of these solutions has been hampered in the past by a lack of accepted standards and the expense of these solutions relative to their performance.
All of these wireless LAN technologies provide "mobility" at OSI Reference Model Layer 2. This allows the benefits of mobility within a highly localized area, defined by the area of coverage for the medium in use. Conventional routing technology could be employed to interconnect these mobile islands. However, as we have seen, the mobility provided by conventional routing technology is limited by static addressing and slow convergence of routing protocols.
Infrared systems can be either point-to-point directed infrared solutions or point-to-multipoint diffused infrared solutions. The directed infrared solutions tend to be limited by the interconnection complexity of line-of-sight requirements and congestion at key nodes; they are appropriate for token ring LAN architectures, where each host is "directly" attached to both an upstream and a downstream station. Diffused infrared solutions tend to be limited by the "interference" of daylight and are appropriate for the shared media nature of Ethernet LANs.
In either case, infrared solutions are physically appropriate only for relatively open areas, such as cubicles; infrared transmission will not penetrate walls. The primary advantages of the infrared systems include freedom from FCC licensing and low cost. Infrared solutions involve interfacing standard computing devices to some form of optical transceiver; as always the API is a key consideration. This technology is likely to continue to be a niche solution strongly supported by its adherents.
Narrowband RF systems, such as Motorola's Altair wireless Ethernet system, are able to penetrate some walls, but require licensing from the Federal Communications Commission (FCC) and tend to be expensive, especially in light of their somewhat limited capacity and coverage. The need for installation of backbone cabling to connect the hubs providing the sometimes limited coverage further reduce the applicability of such solutions.
Altair operates in the 19 GHz RF spectrum and is a closed, proprietary technology providing approximately 3.3 Mbps effective user throughput in a 15 Mbps transmission medium. A derivative product called VistaPoint provides a wireless link for LANs, such as between neighboring buildings at an effective throughput of approximately 6 Mbps. Of course, security becomes a concern whenever the physical barriers of walls are removed as a deterrent to potential network eavesdroppers.
The predominant wireless LAN technology is currently that of spread spectrum RF systems, which operate in the Industrial, Scientific and Medical (ISM) bands at 902-928 MHz, 2.4-2.4835 GHz and 5.725-5.850 GHz. These frequency ranges are unlicensed by the FCC9.1 and so they can be used by many non-LAN applications, such as garage door openers.
There are many vendors of spread spectrum technology. The spread spectrum LAN systems have recently been standardized by the IEEE 802.11 standard9.2 . Typically a low effective user bandwidth solution, this technology provides approximately 1-2 Mbps of usable bandwidth, versus the Ethernet standard supporting transmission rates of 10 Mbps.
There are two forms of spread spectrum technology, both of which spread a baseband signal over a wider range of frequencies in order to achieve resistence to noise and fading and thus gain in performance. It was this resistence to noise (i.e., jamming) which encouraged the development of spread spectrum technology for military applications.
The first spread spectrum technique is called frequency hopping spread spectrum, in which the baseband signal is spread by hopping from carrier frequency to carrier frequency within the ISM band in a pseudorandom fashion; both transmitter and receiver must know the hopping sequence and dwell time at each frequency prior to the transmission.
The second form of spread spectrum is called direct sequence spread spectrum, in which each transmission is modulated by a pseudorandom binary sequence which serves to spread the waveform spectrum; a correlator at the receiver evaluates the energy at the binary sequence-defined frequencies and despreads the signal prior to decoding it.
A recent development is the Metricom Ricochet wireless metropolitan network, which is an outgrowth of the ISM band spread spectrum LAN technology. Metricom, founded in 1985, constructs private wireless networks for utilities for wireless data acquisition and monitoring of public utility grid performance. These private networks were collectively called Utilinet, with a protocol which could be licensed from Metricom for other purposes.
Ricochet is a campus or metropolitan network, which is available in a few locations, primarily in Silicon Valley. Rather than pursuing a "Field of Dreams" strategy of universal coverage, Metricom constructs in the locations where specific customers have already been identified.
The high-level architecture of Ricochet consists of metropolitan networks linked together via public switched services. Within a metropolitan area, Ricochet utilizes a mesh of ISM-band frequency hopping spread spectrum RF repeaters, spaced approximately one mile apart.
Ricochet's many RF repeaters are small low-cost low-power units, mounted on telephone poles and buildings to reduce deployment costs. The maximum transmit power of the radios is one Watt, per ISM band restrictions; this limits reception to the previously-mentioned one mile distance and reduces the in-building penetration of the signal.
The low-power nature of Ricochet also increases the number of repeaters necessary to cover large metropolitan areas-on the order of 100,000 base stations will be required to cover the top 30 markets. This will also increase the network interconnectivity and scaling complexity; smaller cells means more frequent handoffs. Ricochet is a strictly LAN solution whose capability for wide area network services is limited by lack of definition of an internetworking scheme.
Within a metropolitan area, Ricochet has a flat topology, with distributed intelligence capable of supporting both automatic alternate route selection and peer-to-peer communications between mobiles. Network elements include Remote Terminal Units (RTUs), Load Control Transponders (LCTs) and network gateways. Three radio types are used: WAN gates (RS-232-based units at entry-point PCs serving as gateways), network radios (repeaters) and status control radios (which communicate status and control information).
Ricochet frequency hopping is based on a pseudo-random pattern using 240 narrowband channels within the 902-928 MHz band. Radios optimize their frequency hopping by being offset with respect to their neighbors. The same channels are used for both repeater to repeater links and communications between mobile and repeater. Under ideal conditions, the RF technology employed provides a transmission rate of up to 100 Kbps with an effective application rate of up to 38 Kbps.
Ricochet mobility is limited somewhat by the complexities of passing control to the mobile in a distributed system. Mobiles must either be stationary or slow-moving (i.e., less than 10 kph). Small cell sizes limit the speed of communicating mobiles due to handoff failures; the repeaters cannot update their routing tables quickly enough if users move too swiftly. The system design encourages constant connectivity by the mobile device. Direct mobile to mobile communications is supported as long as they are within one mile of one another.
Ricochet repeaters are addressed by latitude and longitude, with color codes used to differentiate between radios; each base station has a global position sensing (GPS) receiver. This is a novel way to unite geographic and network locations for routing purposes. Radios build internal node tables, whose entries include coordinate location, ASCII name, signal strength, unique radio address and frequency hopping algorithm offset.
Packets are routed between radios in the mesh based on latitude and longitude coordinates; this coupled with the dynamics of alternate path decision-making eliminates the need for static routing tables. Each radio also maintains traffic history records, including the number of packets handled and forwarded to each neighbor, the number of retries for each neighbor, the number of hops and time needed for each message, the percentage of successful packet deliveries, etc.
Users can typically expect on the order of 3 or 4 RF hops between meshed base stations before their packets reach a landline access point. This number of hops could increase to 30 or 40 in large cities, depending on the location of backbone points of presence. The backbone is a frame relay network.
The Metricom protocol is reliable. Data packets are sent with the address of the network radio nearest to the end destination. Packets sent are followed by queries for reception status; each packet is separately queried. However, this overhead could effectively throttle the total system capacity, especially considering that the same channels are used for both mobiles and repeater links.
Applications access the system in a prioritized manner. Up to 200 packets can be stored by the radios; packets have time-to-live and number-of-routing-hops parameters to help determine which ones to keep. Old messages are discarded by the radios in the mesh as are those requiring too many RF hops.
The system is self-healing thanks to its distributed intelligence. This helps it recover from outages and control congestion via alternate routing. Alternate paths are selected when the first-choice route is busy or unavailable.