MEDIA ACCESS CONTROL FOR WIRELESS SYSTEMS
The invention relates to media access control (MAC) for wireless systems, particularly
wireless Carrier Sense Multiple Access systems
The invention has particular, though not exclusive, applicability to the Media Access
Control (MAC) protocol, known as Distributed Foundation Wireless MAC
(DFWMAC) drawn up by the IEEE 802.11 Committee, tasked with the role of
standardising protocols for Wireless Local Area Networks (WLANs) - see IEEE,
Draft Standard IEEE 802.11, Wireless LAN, P802.il/Dl, December 1994.
DFWMAC is a hybrid multiple access protocol because the existence of an Access
Point (AP) is not necessary for protocol operation. The protocol also enables an ad
hoc network with distributed control to be established. It is also capable of supporting
both asynchronous data services and delay sensitive applications by dividing
transmission time into frames, each frame consisting of two parts - a 'controlled' part
and a 'random' part. The physical layer uses Spread Spectrum (SS) access; both
Frequency Hopping (FH) and Direct Sequence (DS) SS approaches have been
adopted.
The MAC protocol uses a contention mechanism to allow stations to share a wireless
channel. This mechanism is based on a Carrier Sense Multiple Access (CSMA)
mechanism, similar to the IEEE 802.3 standard. An extension of the CSMA to
include a Collision Detection mechanism (CSMA-CD) is not possible in the radio
environment, because a station cannot, simultaneously, transmit and receive on the
same channel, as required in the part of the CSMA-CD protocol that detects for
collisions (when two or more stations transmit at the same time). Therefore, a
wireless terminal is not always able to determine that a collision has occurred until the
end of the transmission period, making the detection of the collision very inefficient,
since scarce resources are being wasted. However, the IEEE 802.11 MAC extends
the CSMA protocol by introducing a Collision Avoidance mechanism (CSMA-CA),
which reduces the collision probability. The basic protocol used in the CSMA-CA
protocol is known as Distributed Coordination Function (DCF).
DCF is implemented in every station. It is used both for offering asynchronous data
services, and as the basis for the development of a mechanism for offering 'delay
bounded' service to delay sensitive applications. The CSMA-CA protocol is a
variation of the usual CSMA protocol described in "Telecommunications Networks:
protocol, modelling and analysis" by Mischa Schwartz, Addision Wesley series in
Electrical and computer engineering, Addison-Wesley 1987. A node first listens to
the channel to determine if the medium is available, and transmits only when the
medium is idle. 'Collision avoidance' reduces the probability of a collision among
contending terminals by calculating a random idle time at each terminal, during which
the terminal defers transmission, waiting to see if the medium remains idle. A
graphical presentation of this mechanism, is shown in Figure 1(a).
A characteristic of the CSMA-CA protocol is the introduction of various values of
InterFrame Spacing (IFS) to give priorities to different types of MAC packets. For
the basic mechanism now described by way of illustration, only the Distributed IFS
(DIFS) time is important. From the end of a frame transmission, the access procedure
is the following:
All stations do not transmit any information for a time period DIFS.
The stations that have new packets to transmit (terminal A, say) calculate a
random backoff number, which is a timer that indicates when this packet can
be transferred (in time slots). However, the 'countdown' process starts after
the expiration of the DIFS time (that is when the medium becomes idle). This
random backoff time is given in Equation 1, where CW is the value of the
contention window at the time of the calculation, and n a uniformly selected
random number within [0,1].
(backoff time)=CWn (1)
In the illustrated example, the 'countdown process' for terminal A starts at the end of
the DIFS interval (at time t,) but stops (at time t^) when the medium being accessed
becomes busy. The 'countdown process' recommences at the end of the next DIFS
interval (at time t^ for the remaining part of the backoff time (Δt), and terminal A will
start to transmit (at time t4) when the backoff time has expired, provided the medium
remains idle.
The slot duration depends on the physical layer implementation. Its value is chosen
in such a way that stations starting transmissions in different slots are guaranteed not
to collide. The value of the CW in Equation 1 is initialised to CW™n.
Retransmissions cause the value of CW to grow exponentially (binary exponential
backoff) up to CW™. In Figure la it can be seen that only one random backoff
duration is being calculated for each packet. As explained, this duration is used as the
starting value of a timer that only counts down when the medium is idle and stops
when the medium is sensed busy. The packet is transmitted when the timer reaches
zero. For the interest of fairness, a station cannot send packets that are separated only
by a DIFS interval; it must calculate a backoff period, as described above, for each
separate frame, resulting to a Single Station Throughput (SST), where 'packet time'
is the length of the packet in time units (slots):
SST _1 =1 +[DIFS+V2CWmiaV(packet time) (2)
The basic CSMA-CA mechanism for the IEEE 802.11 MAC protocol has been further
enhanced with:
(i) MAC level acknowledgements and retransmissions. The acknowledgements
have priority over the rest of the packets, because the station that issues an
acknowledgement (after the reception of a packet), uses the Short IFS (SIFS,
as shown in Figure 1(a)) timer to decide that the medium is idle.
(ii) Special Ready To Send/Clear to Send (RTS/CTS) frames, which implements
a mechanism for bandwidth reservation, reducing the hidden terminal problem
in ad hoc WLANs. The CTS packet follows the RTS and has increased
priority over the other packets, since it also uses the SIFS timer to decide on
the state of the channel.
(iii) A contention free, access method, known as Point Coordination Function
(PCF), through the use of the Point IFS (PIFS) timer, with a value between
' SIFS and DIFS (see Figure 1(a)). Terminals transmitting in the PCF have
higher priority packets, because they decide that the channel is free faster than
the terminals using the DIFS timer.
The point coordinator (one of the terminals in an ad hoc network, or an AP, if
available) can take control of the frame transmission by using the PIFS to access the
medium. Assuming the existence of an AP, the operation of the PCF is now
explained: The AP informs each terminal of its turn to transmit, by polling it. The
terminal transmits information, receives an acknowledgement and the AP polls the
next terminal in its polling list. Each of the mobile stations can appear more than once
in the polling list, if necessary. The AP initiates the PCF periodically (every
SuperFrame (SF) period). All the terminals in the coverage area of an AP being
informed of the value of this period and the maximum duration of the PCF. At the
end of the PCF period, the AP informs the mobile stations of the beginning of the next
DCF period, by sending a certain packet. If the transmitted packets are being
acknowledged by the receiving station, the acknowledgement transmission takes place
when the SIFS timer expires (SIFS is shorter than PIFS). If a terminal does not react
within a SIFS time interval, then the PCF (the AP) shall resume control and transmit
the next frame after a PIFS gap.
The use of the above two access mechanisms (the PCF and the DCF) divides the SF
period into contention free and contention parts, as shown in Figure 1(b). The lengths
of each period are manageable objects. The contention free period is limited to allow
coexistence between contention and contention free traffic. The maximum time that
is allowed to be allocated to these services in a SF is such that at least one maximum
size MAC Protocol Data Unit (PDU) can still be transmitted during the SF period. A
phenomenon, known as SF stretching effect could take place, because a station may
transmit a long MAC packet, that interferes with the start of the next SF period (as
shown in Figure 1(c)). Since transmissions in the first, contention free, part have
priority over transmissions in the second contention part, different kinds of traffic can
be supported. Generally speaking, the first part (which uses the PCF) can be used for
delay sensitive services, such as voice and video, whereas the second part (which uses
the DCF) can be used for delay independent services, such as general computer data
transmissions.
The transmission rate of the IEEE 802.11 physical layer is either 1 or 2 Mbps, per
Access Point (AP) (the physical layer overhead is always being transmitted at 1
Mbps). Several APs can be included in one location (i.e. a base station), resulting in
a higher bit-rate, able to satisfy most of the ATM services requirements as described,
for example, by Raif O. Onvural in Asynchronous Transfer Mode Networks:
Performance Issues, Artech. House Inc 1994 ISBN - 0-89006-662-0. A simplified
structure of the MAC frame, for data packets only, is shown in Figure 1(d). More
details about the IEEE 802.11 MAC protocol, the different transmitted messages and
the frame structure, can be found in the afore-mentioned IEEE, Draft Standard.
In general, random access protocols like CSMA-CA cannot adequately support delay
sensitive applications. The delay in transmission encountered by a data packet
depends on the offered load of the network. Random access techniques are suitable
for multiplexing bursty sources and have the capability to keep the delay per packet
low provided the aggregate traffic is a small portion of the system capacity. If this is
the case for a WATM LAN, for example, then the DCF can be used and an adhoc
network created amongst terminals. However, if the WATM LAN is required to
operate more efficiently and offer high QoS to delay sensitive applications, even when
the offered load approaches system capacity, then the PCF approach would need to
be used.
An objective of the present invention is to improve efficiency of operation and QoS,
even when the offered load approaches system capacity. To that end, there is
provided in the MAC protocol at least one additional IFS defining an interval
intermediate that of the DIFS and the PIFS.
Accordingly, the present invention provides an Asynchronous Transfer Mode (ATM)
transmission method for use in a wireless communications system, the method
employing a protocol including a Distributed Coordination Function (DCF) having
a Distributed Interframe Spacing (DIFS) and a Point Coordination Function (PCF)
having a Point Interframe Spacing (PIFS), wherein said protocol includes at least one
additional Interframe Spacing (IFS) defining an interval(s) intermediate the interval
defined by said DIFS and the interval defined by said (PIFS).
In this way, the ATM QoS can be mapped to at least three different priorities,
respectively defined by the PIFS (used for real time (delay sensitive) ATM services),
the DIFS and the or each additional IFS (used for non-real time ATM services). This
approach assumes, of course, that the associated primitives are appropriately modified
to be compatible with the new protocol.
The or each additional IFS is capable of creating different priorities amongst different
non-real time services such as nrt VBR (non-real time Variable Bit-Rate), ABR
(Available Bit-Rate), UBR (Unspecified Bit-Rate) and UBR+.
Embodiments of the invention are now described, by way of example only, with
reference to the accompanying drawings: of which
Figure 1(a) illustrates a known CSMA-CA access mechanism,
Figure 1(b) illustrates a MAC protocol Superframe structure,
Figure 1(c) illustrates stretching effect of the Superframe structure of Figure 1(b),
Figure 1(d) illustrates a simplified packet structure according to the IEEE 802.111
MAC protocol and
Figure 2 shows a CSMA-CA access mechanism according to the invention.
At least one additional IFS is provided to create different priorities between non-real
time ATM services that use the DCF or the contention part of the SF period.
Referring to Figure 2, a first additional IFS defining an interval ABRIFS supports
ABR traffic and serves to differentiate ABR traffic from UBR traffic (which is
supported by the DIFS). Differentiation between ABR traffic and other types of ATM
non-real time traffic (such as UBR traffic) is needed when two stations exchange
RTS/CTS messages to reserve resources (band width) for, and transmit, those
different types of traffic (for hidden node problem alleviation).
Provided that PIFS < ABRIFS < DIFS there will be no interference between the
contention free and contention parts of the SF.
In another embodiment, also shown in Figure 2, a second additional IFS defining an
interval nrt-VBRIFS is also provided. This additional IFS supports nrt- VBR traffic
and serves to differentiate the nrt- VBR traffic from ABR traffic (which is supported
by the ABRIFS) and from UBR traffic (which is supported by the DIFS).
The nrt-VBRIFS could be the same as the ABRIFS. In this case, even though the nrt-
VBR and ABR services would have the same access priority (because their IFSs are
the same) they could still be treated differently provided the ABR service is permitted
to transmit larger MAC packets than the nrt-VBR service.
However, if the nrt-VBRIFS and ABRIFS have different values, then interference will
be avoided if PIFS<nrt-VBRIFS<ABRIFS<DIFS. Again, the afore-mentioned
restrictions on packet size may apply to allow "fairness" in the use of the medium .