WiMAX and Its Applications (2)

5 Performance of MAC Layer
The core of the IEEE 802.16 Media Access Control (MAC) layer is the connection-oriented Common Part Sublayer (CPS). The CPS implements system access control, scheduling services, bandwidth request and allocation, connection setup and maintenance, network access and initialization, initial ranging and periodic ranging, etc. To support mobility, IEEE 802.16e introduces other new features into the MAC layer, such as support for handover, sleep mode and idle mode for power saving.

5.1 Network Topology
Worldwide Interoperability for Microwave Access (WiMAX) identifies two topology structures: Point-to-Multipoint (PMP), and optional Mesh, as shown in Figures 5 and 6 respectively. The communications in the PMP mode can only be performed between the Subscriber Station (SS) and the Base Station (BS), while in the Mesh mode SSs can communicate with each other directly. PMP is a centralized topology structure, and BS is the center of the whole system, where all the transmission will start and end. As for the Mesh mode, it is totally different: Each node (such as SS) can set up a communication with other nodes, so BS with multi-hop will have a broader coverage. At present, the MAC layer specifications in IEEE 802.16 are mainly defined for the PMP mode, and most WiMAX systems have been deployed in the PMP mode.

5.2 Media Access Mechanism
All wireless access systems adopting the shared channel mode should take the MAC mechanism into consideration. Different from the Carrier Sense Multiple Access (CSMA)/ Collision Avoidance (CA) policy of IEEE 802.11, the time resources are fragmented for distinguishing uplink and downlink on the physical layer of IEEE 802.16. Each physical frame has a fixed length, consisting of the uplink and downlink parts. The uplink/downlink handoff point can be adjusted through the control of the MAC layer. In the Time Division Duplex (TDD) mode, each frame consists of n  time slots. The downlink in broadcasting is first, and the uplink that is from SS to BS follows. Taking the frame structure of the Orthogonal Frequency Division Multiplexing (OFDM) physical layer as an example, every downlink sub-frame begins with the pilot for synchronizing the physical layer, followed by the Frame Control Header (FCH). The first downlink burst after FCH contains such MAC layer management information as the Downlink Bandwidth Allocation Map (DL-MAP), Uplink Bandwidth Allocation Map (UL-MAP), Downlink Channel Descriptor (DCD), and Uplink Channel Descriptor (UCD). The information describes which SS the following downlink data belongs to, and which modulation and coding mode are used, as well as how to allocate the uplink time on the local frame. Thus, once detecting the head of a frame, SS will know which mode is used and when to receive and send data. This design not only eliminates the competition on the uplink, but also schedules and allocates all the system recourses on BS in a centralized mode.

     For he broadband wireless access system, this media access mechanism is flexible and fair. Firstly, each SS has a chance to send data, which avoids some SSs failing to get channels for a long time. Secondly, each SS sends data within its own sending period, which ensures that there is only one data stream transmitting on the media at any moment. Finally, the mechanism is convenient for the control of Quality of Service (QoS), service priorities and bandwidth.

5.3 QoS Assurance Mechanism
WiMAX is the first wireless access standard that provides QoS assurance on the MAC layer. It is well known that such factors as multi-path and attenuation on wireless channels may result in a high error bit rate and a high packet loss rate. Therefore, the reliability and effectiveness of data transmission is difficult to be ensured. To meet the high requirements of the high-speed multimedia services for time delay, bandwidth and loss rates, WiMAX defines a series of strict QoS control mechanisms in the MAC layer, so it can provide different services with different service quality levels in the wireless access network. In addition, the mechanisms are connection-oriented.

     QoS requirements of WiMAX include the following aspects:

• Configuration and registration function for pre-configuring SS-based QoS service flows and traffic parameters;
• Signaling function for dynamically establishing the service flows and traffic parameters supporting QoS;
• Utilization of MAC scheduling and QoS traffic parameters for uplink service flows;
• Utilization of QoS traffic parameters for downlink service flows;
• Grouping of service flow attributes into named Service Classes, so thatupper-layer entities and external applications may request service flows with specific QoS parameters in a unified mode.

5.3.1 Connection and Service Flow
Connection is a basic concept in WiMAX. It defines a group of one-way packet data that is being transmitted. The service-specific Convergence Sublayer (CS) classifies the data packet from the upper layer first, and then maps it to a specific connection according to certain rules. Afterwards, the operation and scheduling of the data is performed in the unit of connection. The connection itself represents the idea of QoS. The MAC layer can set different QoS parameters aiming at each connection, including bandwidth, time delay and delay jittering, and further builds a special QoS assurance mechanism for this connection. After entering the CS classifier, the data with different priorities or QoS requirements are allocated to different connections.
Each connection has a unique 16-bit Connection Identifier (CID). BS manages all connections in a cell, and 16-bit CIDs decide that the maximum number of connections is 65,536. Either BS or SS can initiate an SS connection, and they can also modify or delete a service connection. When SS accesses the network and is initiated, the basic, primary and secondary management connections are set up between BS and SS. The three connections are used to transmit the management services with different QoS levels: The basic management connection is to transmit short and real-time control messages; the primary management connection is to transmit long control messages that can tolerate more time delay; and the secondary connection is to transmit the control messages that are insensitive to time delay.

     Besides distinguishing services with different priorities, the connection plays an important role in addressing in an IEEE 802.16 network. Every BS or SS has one 48-bit MAC address, which is only used to set up the management connection in the initial ranging, and to make inter-authentication between BS and SS during the authentication process. Next, the data transmission is identified by the CID in the MAC header. This uniform addressing mode can relieve the MAC layer of the management burden. Moreover, it can reduce the transmission overhead of services. For example, in Voice over IP (VoIP), utilization of CID can effectively compress the header.

     A service flow is a one-way packet transmission service provided by the MAC layer. It can be either on the uplink or on the downlink. A service flow takes a set of QoS parameters as the basic characteristics, including time delay, delay jitter, throughput, etc. BS also manages all service flows, besides connections.
A service flow in the IEEE 802.16 system falls into three types: provisioned; admitted, and active. The provisioned service flow is configured by the upper layer application programs or external system in advance but it does not bear the actual service data. The admitted service flow is an intermediate status when the BS has allocated resources to this service flow but the preset parameters are not activated. Only after the service flow is activated by BS or SS, it has the resources committed by the BS, and the communication can run normally.

     All service flows contain one 32-bit Service Flow Identifier (SFID). Besides, the accepted or activated service flow contains one 16-bit CID, which maps the service flow onto the connection that the CID uniquely identifies. The basic principle for WiMAX to provide QoS is to associate the packet data transmitted via the MAC interface to the service flow with a specific CID, while the service flow is beared by a connection. SS and BS provide the connection with the corresponding QoS according to the QoS parameter set defined by the service flow.

     Therefore, service flow and connection are two concepts closely associated. A service flow indicates the service type that the MAC layer provides for the upper-layer service data, and is used to distinguish QoS requirements for different services. A connection is used for the MAC layer to manage and schedule the data internally, in order to meet the specific QoS requirements. Through the mapping of service flow with connection, the requirements connect with the implementation.

5.3.2 Scheduling Service
The so-called transmission scheduling is to select data to send in a specific frame/allocated bandwidth. It is implemented by the BS on the downlink and by the SS on the uplink. The factors involving the scheduling include the scheduling service type assigned by the service flow, QoS parameter value of the service flow, data to be sent, size of authorized bandwidth etc. The scheduling service indicates the data processing mechanism that MAC scheduler supports. Each type of scheduling services is defined by one group of QoS parameters. These parameters determine the quality of the data transmission in all aspects. Each connection is associated with a specific scheduling service, which equips the data packet with the corresponding QoS requirements before the scheduling.

     With respect to service flows on the uplink, the IEEE 802.16d identifies four types of the scheduling services. They are Unsolicited Grant Service (UGS), real-time Polling Service (rtPS), non-real-time Polling Service (nrtPS), and Best Effort (BE) service.

     (1) UGS
     UGS serves to transmit the real-time service flow of fixed-length data packet on a periodic basis, such as the T1/E1 and VoIP service without silence suppression. In this service, the BS assigns the SS with the fixed bandwidth periodically. This bandwidth is enough to hold the fixed-length data associated with the service flow. The UGS mode avoids the overhead and time delay generated from the SS’s bandwidth requests, and ensures the real-time requirement of the service flow.

     The mandatory QoS parameters for the UGS includes the Maximum Sustained Traffic Rate, the Maximum Latency, the Tolerated Jitter, and the Request/Transmission Policy.

     (2) rtPS
     rtPS serves to transmit the real-time service flow of the variable-length data packet on a periodic basis, such as the Moving Picture Experts Group (MPEG) video streams. It provides the real-time and periodical unicast request opportunities to meet the real-time requirements of the service flow. At the same time, it allows the subscribers to designate the size of the required bandwidth. Compared with UGS, rtPS asks for larger request overhead, but it supports the changeable authorized bandwidth to optimize the efficiency of data transmission.

     The mandatory QoS parameters for the rtPS include the Minimum Reserved Traffic Rate, the Maximum Sustained Traffic Rate, the Maximum Latency, and the Request/Transmission Policy.

     (3) nrtPS
     nrtPs serves to transmit the service flow that asks for the minimum data rate, has data packet of variable length, and is insensitive to the time delay, such as the File Transfer Protocol (FTP) service. This service provides timely unicast polling opportunities to ensure the service flow to acquire bandwidth request opportunities even when the network is congested. The BS usually polls all nrtPS connections every several seconds or shorter time.

     The mandatory QoS parameters for the nrtPS include the Minimum Reserved Traffic Rate, the Maximum Sustained Traffic Rate, the Traffic Priority, and the Request/Transmission Policy.

     (4) BE Service
     The BE service serves to transmit the service flow that does not need minimum reserved data rate, such as the Internet service. This service works only in idle cases.
The mandatory QoS parameters for the BE include the Maximum Sustained Traffic Rate, the Traffic Priority, and the Request/Transmission Policy.

     To improve the bandwidth utilization, the 802.16e defines the Extended real-time Polling Service (ErtPS), which supports the services with silence suppression, such as VoIP. As in the UGS mode, the BS in the ErtPS mode automatically allocates the bandwidth to the SS. The bandwidth in UGS is fixed, while in ErtPS it is dynamic and changeable.

5.3.3 Bandwidth Allocation and Request Mechanisms
The bandwidth of the UGS connection is fixed, while all other services may need the bandwidth increased or reduced during the connection. Therefore, different from the fixed channel allocation of traditional mobile communications networks, the IEEE 802.16 system uses Demand Assigned Multiple Access (DAMA). When there is a demand from a connection, the system allocates the bandwidth dynamically, to meet the requirements of the QoS and to increase the efficiency of the resources utilization.

     According to the protocol, SS can apply for the bandwidth in the following methods:
     (1) Requests
     Request mechanism allows the SS to directly apply uplink bandwidth from the BS. The request information can be sent in an independent Bandwidth Request (BR), or piggy backed during the data transmission. In the former case, the BR can be incremental or aggregate, while it can only be incremental in the latter.
SS will request for the aggregate bandwidth once after accessing the network, re-entering into the network from the idle mode, handover, or before using the incremental BR. Meanwhile, the self-correction nature of the request/grant requires the SS to send the aggregate BR periodically according to the QoS and the link quality.

     (2) Grants
     BR that the SS sends is based on a single connection. However, the bandwidth grant is addressed to the basic CID of the SS, instead of the CID of the connection. It is impossible to judge which request has been granted, so when the granted bandwidth is smaller than the expected value, SS will decide, according to the status of the request and the latest information received from the BS, to backoff and request again or discard the Service Data Unit (SDU).

     (3) Polling
     Polling refers to the process that the BS allocates a specific bandwidth to SS for sending BR. If the resources are abundant, the allocation may be made to each inactive SS individually, which is called unicast polling. Any SS with the active UGS connection can also apply to join the unicast polling by setting Poll-Me (PM) bit in the MAC header of the UGS connection. If the resources are inadequate, multicast or broadcast may be used to allocate a common bandwidth resource to one group of SSs for sending BR. To decrease the collision, only the SS that require the BR will give a response and decide in which TS to send the initial BR according to the contention resolution algorithm prescribed in the protocol.

     (4) Contention-Based Focused BR
     This mechanism is used in the OFDM physical layer. SS can send the focused request message in the Registration (REG) domain of the data frame. The mechanism is implemented by selecting one Contention Code out of the eight possible codes with equal probability, and transmitting the selected on the Contention Channel consisting of four sub-carriers. After detecting the message, the BS allocates a specific bandwidth to this SS for sending BR.

     (5) Contention-Based CDMA BR
     This mechanism is used in the Orthogonal Frequency Division Multiplexing Access (OFDMA) physical layer. OFDMA defines one Ranging Subchannel and three pseudo-random Ranging code subsets for BR, initial ranging and periodic ranging respectively. When applying for bandwidth, an SS can randomly select one CDMA Ranging code from the Ranging code subset for BR, and then send it through the Ranging Subchannel. After detecting the message, the BS allocates a specific bandwidth to this SS for its sending BR.

5.3.4 Automatic Repeat Request
Automatic Repeat Request (ARQ) is a QoS-related function. It is optional in the IEEE 802.16, but not applicable to the physical layer adopting single-carrier technology. Different from the Hybrid Automatic Repeat Request (H-ARQ) in the physical layer, ARQ belongs to the MAC layer. H-ARQ can only be enabled on a per-terminal basis, while ARQ is connection-oriented. During the setup of a connection, the use of ARQ shall be specified and negotiated. One connection cannot contain both ARQ and non-ARQ traffic at the same time.

     The protocol defines the size of an ARQ block. If the connection adopting ARQ supports MAC SDU fragmentation, the size of each fragment shall be the integral times of the ARQ block. The sequence number of the first ARQ block in this fragment is recorded in the sub-frame header. The receiver can detect the possible errors in the receiving data with CRC-32. If there are errors, the receiver can find the fragment with errors according to the ARQ sequence number. Afterward, the ARQ mechanism will start retransmission on a per-fragment basis.
The ARQ feedback information can be sent on the SS basic management connection as independent MAC management information, or piggy backed in data transmission of the connection.      (to be continued)

[Abstract] Worldwide Interoperability for Microwave Access (WiMAX), an emerging broadband radio access technology, has attracted much attention from the whole telecom industry in recent years. Its main features are the high-speed transmission rate, large coverage, support for mobility, QoS guarantee and all-IP architecture. The technology fulfills the integration of packetized data, broadband access and mobilized terminal; therefore, it has a bright future for wide application. This lecture discusses WiMAX in four parts, and this part introduces the MAC layer and its QoS mechanism.