|A new IEEE standard promises to bring interoperability to the wireless LAN marketplace
The wireless LAN marketplace is heating up. By giving you the ability to roam throughout a coverage area while remaining connected to your LAN-based services, wireless technology is a natural fit for today's horizontal corporations and mobile workforce. But as the market is flooded by interested parties--hardware manufacturers, system integrators, and computer manufacturers--the need grows for interoperability between competing products.
The worldwide authority on standardization in wireless LANs is the Wireless Local-Area Networks Standards Working Group, IEEE Project 802.11. Since 1990, the Project 802.11 committee has worked to establish a universal standard fo r the wireless marketplace. Recently, the committee selected the DFWMAC (distributed foundation wireless media access control) protocol proposal from AT&T Global Information Solutions/NCR Microelectronic Products Division's Wireless Communications and Networking Division, Symbol Technologies, and Xircom as the foundation for the development of a wireless LAN standard. Widespread adoption of the technology embodied in DFWMAC will ensure a vital and interoperable wireless LAN marketplace.
Requirements and Architecture
The initial task of the Project 802.11 committee was to analyze the applications and environments in which wireless networks are used. As early as March 1992, the committee formally established the functional requirements for a wireless LAN protocol.
The Project 802.11 committee established the minimum functional bandwidth at 1 Mbps. This was deemed necessary for common operations such as file transfer, program loading, transaction processing, multimedia, and manufacturing process control. For applications such as digital voice and process control, which require transmission of real-time data, the committee decided to require support for time-bounded services, which limit the packet delay variance. It also identified the need for reliable operation in a wide range of environments, including financial, retail, office, school, and industrial settings. In addition, it was decided that mobile computing should at least support pedestrian speeds of several miles per hour, with a vehicular-speed option for industrial users.
To address these requirements, the Project 802.11 committee formulated a basic architecture for wireless LAN systems. Generally speaking, wireless networks break into two types. The first type is infrastructure-based networks that let you roam through a building (e.g., a store, a hospital, or a manufacturing floor) while maintaining a connection with the organization's computer resources. Usually, wired networks form the foundation for wireless-network infrastructu res. The second type is an ad hoc network that any number of users can set up instantly, as might be desired when meeting in a conference room, for example. The 802.11 architecture allows for overlap by using the same basic access protocol for both ad hoc and infrastructure-based networks. The basic architecture lets multiple networks share the same medium, using the same channel, thus ensuring a high degree of efficiency in frequency usage.
The Project 802.11 committee also defined the different components of a wireless LAN. A single cell within an infrastructure-based network is called a Basic Service Area, or BSA. The size of any cell is dependent on the environment and the power of the wireless transceivers. Any single BSA can contain a number of discrete groups of wireless stations. Multiple BSAs can cover larger areas, interconnected by APs (Access Points) and a distribution system (which is usually wired). Such interconnected BSAs form an ESA (Extended Service Area). The group of stations that a re associated to the same AP is called the Basic Service Set, or BSS. The set of stations within multiple BSSes that are connected via a distribution system forms an ESS (Extended Service Set). The figure "Architecture for Wireless LANs" illustrates the basic architecture of wireless LANs.
The Reference Model
The Project 802.11 committee uses a reference model that divides protocols for wireless communications into two main groups. The first group of protocols is a common MAC (media access control) specification for all wireless networks. A single, medium-independent MAC protocol provides a unified network interface between different wireless and wired networks.
The second group of protocols are the PHY (physical) specifications for medium-dependent protocols. In wireless communications, the medium is defined by signal characteristics in a particular bandwidth of frequencies. There are different PHY specifications for each frequency bandwidth supported in Project 802.11. For example, there ar e different PHY specifications for the 915-MHz bandwidth, the 2.4- and 5.2-GHz bandwidths, the infrared bandwidth, and so on. The figure "One MAC for All" shows PHY layers supported by the IEEE Project 802.11 proposal.
Defining the reference model resulted in a list of criteria that any MAC proposal had to address, with support for multiple PHY specifications being one of the most important. Because NCR, Symbol Technologies, and Xircom individually use different PHY layers, support for multiple PHY technologies was built into the DFWMAC protocol proposal. Other proposals on the table were more focused toward a specific PHY layer. Other important criteria that must be met include power management and time-bounded services.
The lowest protocol level in DFWMAC is the DCF (Distributed Coordination Function), which supports asynchronous communication between multiple stations. DCF supplies the basic medium access that allows for automatic medium sharing between similar and dissimila r systems. Contention between multiple stations wishing to access the same medium is resolved through a mechanism called CSMA/CA (carrier-sense multiple access/collision avoidance) with acknowledgment.
The CSMA function in DFWMAC is similar to the one Ethernet uses. The carrier-sense mechanism determines whether the signal energy in a particular frequency bandwidth is above a certain threshold. If the signal strength is below the threshold, that frequency bandwidth is available for wireless data communications, and the transmitter sends a parcel of data called a frame. If the signal strength is above the threshold, the medium is considered busy. When the carrier-sense mechanism determines that such multiple accesses to the medium are occurring (i.e., that the medium is busy), the transmitter waits for a short while before trying to retransmit. This waiting period is called backoff. To reduce the probability of access collisions and provide fair access to the medium by all stations, the time gap between stations accessing the medium is varied by backoff periods of random lengths.
DFWMAC's CSMA/CA and Ethernet's CSMA/CD (carrier-sense multiple access/collision detection) use a carrier-sense mechanism to determine whether other transmitters are using the medium. In both cases, if the transmitter senses the medium is free, then the transmission of the frame starts immediately. However, if the transmitter senses the medium is busy, CSMA/CD and CSMA/CA use slightly different methods to resolve the contention.
In Ethernet, when the transmitter detects a busy medium, it defers access until the end of the current frame plus an IFS (interframe space), or silence period. It then transmits its frame. If more than one station is deferring simultaneously, they will possibly start transmitting at the same time, causing a collision. Note that the probability of a collision occurring is highest when the medium becomes free. Collision frequency also depends on the total network load.
In Ethernet, the CD (collision detection) function detects this collision: All colliding transmitters sense collisions and generate a random backoff delay. After the delay, the transmitters reexamine the medium again to see if it has become free. By contrast, in wireless networks, the CD function is not viable. This is because the dynamic range of the signals on the medium used by multiple networks or stations is very large. As a result, no CD techniques can be used to resolve access contention. As collisions cannot be detected, the likelihood of transmission failures increases; therefore, lost frames in a wireless CSMA-only implementation become more likely.
To keep collisions from destroying data, DFWMAC uses a CA mechanism. Backing off and waiting until the medium is free is not sufficient to provide reliable communications, so the CA (collision detection) mechanism adds a MAC-level acknowledgment to ensure the integrity of individual packets.
The acknowledgment protocol allows for data recovery at a low level, which is essential in a wireless environment to resolve reliability problems arising from access collisions and interference. To allow detection of a lost frame due to interference or collisions, the destination station (or stations) returns an acknowledgment immediately following a successfully received frame. When the acknowledgment fails to return, a MAC transmitter can recover from this error by retransmitting the frame after a random retransmission backoff.
DFWMAC defines an efficient backoff algorithm that is stable at high loads. The algorithm uses exponential backoff for retransmissions. To support coexisting asynchronous and time-bounded services, the algorithm also supports different priority levels that different IFS delays control. The figure "Avoiding Collisions" shows the basic access mechanism. The key procedure is that a station that wants access to the medium needs to sense the medium first to ensure that a minimum IFS has been assigned.
The protocol defines three backoff prior ities. The exact backoff period varies randomly. The priorities are as follows:
-- SIFS (short IFS). This, the shortest IFS, is used for all immediate response
actions. These actions include acknowledgment frames, RTS (request-to-send)
frames followed by CTS (clear-to-send) frames, and any contention-free
response frame sent during time-bounded services.
-- PIFS (point coordination function IFS). A mid-length IFS, this is used for
station polling for time-bounded services.
-- DIFS (distributed coordination function IFS). The longest IFS, this is used
as a minimum delay for asynchronous frames in the contention period.
In general, using a DCF-based CSMA/CA access technique extended with a MAC-level acknowledgment protocol ensures the reliability of transmitted wireless frames and the efficient recovery of transmission failures at the MAC layer itself.
The mobile nodes used for wireless transmission are usually small, hand-held , battery-operated devices. Power conservation management functions in DFWMAC allow efficient battery operation while maintaining connectivity and network throughput.
Current wireless-network protocols assume that nodes are always ready to receive frames from the network. With power management, station receivers can be turned off most of the time, saving battery power without affecting functionality. The DFWMAC includes a protocol that lets you switch mobile computers from full-power (running) mode to low-power or sleep mode, where special mechanisms ensure delivery of all wireless data communications. The power conservation provisions are provided in both infrastructure and ad hoc modes. With these mechanisms, battery life in infrequently used scanners or palmtop devices can last for months.
In DFWMAC, time-bounded services are available via an optional PCF (Point Coordination Function). The PCF runs on top of the basic-access protocol to ensure coexistence of both time -bounded and nontime sensitive applications (see the figure "Two-Headed Access").
The PCF uses a superframe concept to ensure contention-free service. Within a given superframe period, the PCF is active in the contention-free period, while the DCF is active in the contention period. The contention-free period can be variable in length on a per-superframe basis without incurring any additional overhead (see the figure "The Superframe").
At the beginning of the superframe, if the transmitter senses that the medium is free, the PCF gains control over the medium. If the transmitter senses that the medium is busy, then the PCF defers until the end of the frame; but when the transmitter gains control over the medium, the data channel is available for the PCF period. In the DCF, a frame can still start near the end of the DCF period. This causes stretching of the superframe, which, in turn, causes the contention-free period to start at variable intervals.
Asynchronous data traffic using DCF auto matically defers until after the contention-free period. This is because the PCF uses the PCF priority level of the CSMA/CA access protocol, which causes a burst of traffic with interframe gaps that are smaller than the minimum DIFS period that the CSMA/CA needs. As DCF is the basic access scheme for both asynchronous and time-bounded services, with DCF running under PCF for time-bounded services, both asynchronous and time-bounded traffic defer to each other when appropriate.
Finalizing the Standard
Ultimately, the DFWMAC protocol proposes to solve the wireless standard debate by providing a complete wireless LAN system that would accommodate both the ad hoc wireless LAN environment and an infrastructure wireless LAN. By accommodating both environments with the same access protocol, DFWMAC bridges the gap in protocol interoperability. The breakthrough Project 802.11 MAC protocol vote, in particular, will bring computing and data communications solutions together to provide people access to informat ion and to each other--anytime, anywhere.