MOBILE WIMAX END-TO-END ARCHITECTURE

IEEE 802.16 standards family designed the physical and the MAC layers to provide broadband services access at the metropolitan scale for fixed, nomadic, portable, and mobile subscribers. The WiAMX forum, whose mission is to promote the interoperability of the broadband wireless access equipments implementing the IEEE 802.16 and ETSI HIPERMAN standards, has created a Network Working Group and a Service Provider Working Group to address higher-layer specifications such as intervendor inter-network interoperability for roaming, multivendor access networks, and intercompany billing. The result of these standardization efforts is the mobile WiMAX end-to-end architecture. The architecture is based on an all-IP platform implementing packet switching; it defines an access network and a common core network while decoupling the access from connectivity IP service. Moreover, the end-to-end architecture is designed to support loosely coupled interworking with existing wireless networks such as UMTS and existing wired networks such as DSL. Besides, a global roaming across WiAMX operator networks is achieved through the support for credential reuse, common billing and settlement, and consistent use of authentication, authorization, and accounting (AAA) services. The WiMAX forum designed the WiMAX network reference model (NRM), which aims at achieving interoperability through identifying the functional entities and reference points and the corresponding communication protocols and data plane treatment within a logical representation of the network architecture. As depicted in Figure 1, the MS, the access service network (ASN), and the connectivity service network (CSN) represent a set of functional entities that may be implemented by a single physical device or by different physical devices. The ASN is formed by at least one BS and one ASN gateway (ASN GW); it implements the access services and represents a boundary for functional interoperability with WiMAX clients and WiAMX connectivity service functions. For instance, the BS manages the MSs in its coverage while the ASN GW relays data to the CSN. On the other hand, the CSN may be defined as a network of Internet gateways, user databases, routers, servers, and proxies providing IP connectivity services to WiMAX subscribers.

Figure 1: The WiMAX NRM.

Mobile WiMAX end-to-end architecture is based on a security framework which provides basic security services and particularly authentication and confidentiality. In fact, each MS authenticates itself to the WiMAX network while the WiMAX network authenticates itself to the MS by implementing consistent and extensible authentication mechanisms. Besides, data confidentiality and integrity, replay protection and nonrepudiation services are guaranteed using applicable key length. Last but not least, MSs have the possibility to initiate and terminate specific security mechanisms such as virtual private networks (VPNs). The end-to-end architecture supports advanced IPv4-or IPv6-based mobility management mechanisms. For instance, vertical handovers can occur with wireless LANs or third generation wireless networks while roaming between different network service providers (NSPs) is supported. Seamless handover is also supported at up to vehicular speed while optimizing the overall network resources through the implementation of dynamic and static home address configurations, dynamic assignment of the home agent in the service provider network, and in the home IP network based on policies, etc. WiMAX mobility management will be further detailed.

MAC LAYER OVERVIEW | MOBILE WiMAX

The WiMAX forum has designed the physical and MAC layers of the mobile WiMAX based on the amendments of the IEEE 802.16e to offer broadband services including voice, data, and video at the metropolitan scale for mobile users. The mobile WiMAX network comprises BSs and subscriber stations (SSs). Each SS is assigned a 48-bit MAC universal address that it used to uniquely identify it toward a BS. Mobile WiMAX uses UL and DL maps to prevent collisions. More specifically, SSs implement the time division multiple access (TDMA) to share the UL while BSs use the time division multiplexing. 

 

UL and DL schedules are exchanged between the BS and the managed SSs in every frame using the UL-MAP and the DL-MAP messages. The MAC layer is connection-oriented; besides, all data communication is associated with a connection. Each connection with its QoS parameters forms a service flow and is identified by a 16-bit connection identifier (CID). MAC layer connections can be compared to TCP connections. 

 

In fact, thanks to TCP, a computer may have simultaneously different active connections for different applications using different ports. With MAC connections, an SS may have many connections to a BS for different services such as network management or user data transport; every connection is characterized by its own bandwidth, security, and priority parameters. When a new SS joins the network, the managing BS assigns to it three CIDs with different QoS requirements used by different management levels which are the basic, the primary, and the secondary management connections. The basic connection enables the transfer of short, time-critical MAC and radio link control messages; the primary management connection is used to transfer larger but more delay-tolerant messages while the secondary management connection is used to transfer standards-based management messages such as DHCP ones. Note that a CID may carry traffic for many different higher-layer sessions. The mobile WiMAX version adapts dynamic modulation and forward error codes to correctly serve the SSs located far from the BS in the rural areas and resists the weather conditions. 

 

Both the BS and the SS can adapt the transmission of the burst profiles by lowering the bandwidth for higher robustness. For instance, the BS always begins by adopting the most robust modulation and forwards error code scheme so that all SSs in the coverage area can correctly receive the DL-MAP and the UL-MAP messages. Meanwhile, an SS may ask its BS to have a longer UL window when needed. The BS and the managed SSs exchange MAC protocol data units (PDUs) carrying MAC management messages or convergence sublayer MAC service data units (MAC SDUs). The MAC PDU has a fixed MAC header, a variable length payload and a cyclic redundancy check (CRC) field. The MAC header may be a generic MAC header (GMH) or a bandwidth request header. The GMH is used to transfer the standard MAC management messages while the bandwidth request header is a header sent without payload to request additional bandwidth. 

 

Both the payload and the CRC fields are optional. It is worth noticing that the MAC header and the MAC management messages are never encrypted to facilitate registration, ranging, and normal operation of the MAC sublayer; however, this decision has opened the door to eavesdropping and other serious attacks. An SS which enters the network may be programmed to register with a certain BS; but generally speaking, it begins by scanning its frequency to detect an operating channel. Scanning consists in listening to each possible frequency until the frame preamble is heard. After detecting that channel, the SS tries to synchronize to the DL transmission by waiting for the DL-MAP stating the map of the timeslot locations in use for the frame. After that, the MS waits for the downlink channel descriptor (DCD) and the uplink channel descriptor (UCD) messages that are periodically broadcasted to specify the modulation and the FEC schemes used on the carrier. After gathering the required information describing the parameters needed for initial ranging transmission, the SS scans the UL-MAP to find an opportunity to perform the ranging. Initial ranging is used to determine the transmit power requirements of the MS to reach the BS. It is worth noticing that each SS should be informed about when to send the ranging request as many SSs may try to join simultaneously the network and highly affect the networks efficiency. Therefore, each new SS will send a ranging request (RNG-REQ) message and wait for the corresponding ranging response (RNG-RSP) message, indicating the timing advance, the power adjustment, and the basic and the primary management CIDs. After correctly determining the timing advance of SS transmissions, the SS and the BS will continue exchanging RNG-REQ and RNG-RSP messages until an acceptable radio link is established. Then, the SS should perform the authentication and the registration processes to enter the network. First, the BS asks the SS for strong authentication. Upon successful authentication, the SS will be able to register to the network. The SS will then establish a secondary management CID to receive secondary messages for different services. For instance, the SS will get an IP address through DHCP and a Trivial File Transfer Protocol (TFTP) address to request configuration files when needed.

QoS is guaranteed thanks to five QoS classes implementing different scheduling mechanisms. Those classes are the unsolicited grant service (UGS), the real-time polling service (rtPS), the extended rtPS (ErtPS), the nonreal-time polling service (nrtPS), and the best effort (BE). UGS fulfils the requirements of real-time communications occurring at periodic intervals such as voice over IP (VoIP); it is given a grant by BS that accommodates the maximum sustained traffic rate of such applications. rtPS uses unicast polling to support periodic real-time variable-size transmissions such as MPEG video streams, but generates more overhead than UGS. ErtPS is a combination of UGS and rtPS; in fact, the bandwidth is allocated without solicitation, but the allocation is done in a dynamic fashion. ErtPS fits well in the case of voice with activity detection like VoIP with silence suppression applications. nrtPS uses unicast polling regularly and allows contention requests to guarantee a minimum data rate for delay-tolerant applications, which generate variable-sized traffic such as FTP. Finally, the BE class guarantees a minimum QoS level for the associated traffic. Therefore, the SSs are never polled individually but are rather permitted to use contention requests and unicast requests.

PHYSICAL LAYER OVERVIEW | MOBILE WiMAX

802.16e WiMAX system profiles will cover 5, 7, 8.75, and 10MHz channel bandwidths for licensed spectrum allocations in the 2.3, 2.5, 3.3, and 3.5 GHz frequency bands. Targeting a worldwide coverage, the WiMAX forum intends to promote the allocation of frequency bands inferior to 6 GHz for civilian applications while considering the available spectrum all over the world. For instance, the 3.5 GHz frequency band is already assigned to fixed services in many countries, the 2.3 GHz band was reserved for the deployment of the WiBro solution in South Korea, while the 2.5 and 2.7 GHz bands have been assigned by the United States for fixed and mobile WiMAX deployment. 

 

Considering the modulation technique, mobile WiMAX is based on the orthogonal frequency division multiple access (OFDMA) technique and particularly on the SOFDMA variant. The frequency division multiplexing (FDM) principle consists in using multiple frequencies to transmit different signals in parallel. This is achieved by assigning a frequency range or subcarrier to each signal then modulating it by data. The multiple subcarriers used for transmitting different signals should be separated by guard bands to prevent interferences. 

 

Orthogonal FDM (OFDM) eliminates the guard bands by using overlapping subcarriers that are spaced apart at precise frequencies; thus achieving orthogonality. With OFDM, the center of the modulated carrier coincides with the edge of the adjacent carrier so that the independent demodulators performing a discrete Fourier transform see only their own frequencies. Redundancy may be guaranteed by scattering some bits over some sets of distant subcarriers. The OFDMA is a multiple access and multiplexing scheme that multiplexes data streams generated by multiple users onto the downlink (DL) subchannels and fulfils uplink (UL) multiple access by means of UL subchannels. The OFDMA symbol structure is made up of data subcarriers for data transmission, pilot subcarriers for estimation and synchronisation purposes, and null subcarriers used for guard bands and DC carriers. Data and pilot subcarriers are organized in a subset of subcarriers called subchannels. The minimum frequency–time resource unit of subchannelization is one slot equal to 48 data tones. 

 

It is worth noticing that there exist two types of subcarrier permutations for subchannelization which are the diversity permutation and the contiguous permutation. The diversity permutation organizes the subcarriers in a pseudorandom fashion to form a subchannel while the contiguous permutation organizes a block of contiguous subcarriers forming a bin. The bin is formed by nine contiguous subcarriers in a symbol with eight assigned for data, and one assigned for a pilot. Generally speaking, diversity subcarriers permutation achieves a high performance with the mobile applications while the contiguous subcarrier permutation fits well to fixed, portable, or low mobility environments. 

 

The SOFDMA technology optimizes the mobile access by assigning a set of subcarriers to particular users. For instance, subcarriers 1, 3, and 7 may be assigned to user 1 while subchannels 2, 5, and 9 to user 2 and so on. Users close to the base station (BS) will benefit from a larger throughput by getting an important number of subchannels with a high modulation scheme. SOFDMA offers multiple bandwidths to allow multiple spectrum allocation and fulfil different usage-model requirements. In fact, the scalability is guaranteed by adjusting the fast Fourier transform (FFT) size with respect to the available bandwidth while fixing the subcarrier spacing at 10.94 kHz. As the resource unit-subcarrier bandwidth and the symbol duration are fixed, scaling bandwidth will not affect higher layers. 

 

The first release of mobile WiMAX will adopt a system channel bandwidth of, respectively, 5 and 10 MHz, with a sampling frequency of, respectively, 5.6 and 11.2 MHz. The FFT size will be either 512 or 1024 while the number of subchannels will be either 8 or 16 and the useful symbol time will be fixed to 91.4 ms. The OFDMA symbol duration is 102.9 ms while the number of OFDMA symbol within a 5 ms frame is 48. The modulation scheme will gradually vary from 16 quadrature amplitude modulation (QAM) to quaternary phase shift keying (QPSK) (four channels) and even binary phase shift keying (BPSK) (two channels) at longer ranges while the power allotted to each channel will be increased. The adaptation of multiple-input multiple-output (MIMO) antenna technique along with advanced coding and modulation achieve peak DL data rates up to 63 Mbps per sector and peak UL data rates up to 28 Mbps per sector in a 10 MHz channel.

 

IEEE 802.16e supports both time division duplex (TDD) and frequency division duplex (FDD) duplexing modes. TDD is adopted when the license-exempt spectrum is used because a unique channel is shared between the UL and the DL traffics which occupy different time slots. Contrarily to TDD, FDD operates with two channels, one is dedicated to the UL traffic and the second is reserved to the DL traffic. Mobile WiMAX in its first release supports only the TDD duplexing mode although the WiMAX forum intends to address particular market opportunities by supporting FDD in future releases. With TDD, it becomes feasible to adjust the DL/UL ratio with respect to the nature of the ongoing traffic so that the DL/UL asymmetric traffic is efficiently supported. Besides, TDD guarantees channel reciprocity for better support of link adaptation, MIMO, and other advanced antenna technologies. Meanwhile, TDD offers a greater flexibility for adaptation to varied spectrum allocations as it uses the same channel for both UL and DL traffics. Last but not least, transceivers designed for TDD implementations are less complex and less expensive. An OFDM frame structure for a TDD implementation is divided into DL and UL subframes separated by transmit/receive and receive/transmit transition gaps (TTG and RTG) to eliminate DL and UL transmission collisions. The frame control information is carried by the preamble, the frame control header (FCH), the DL-MAP, and UL-MAP, the UL ranging, and the UL fast channel feedback (UL CQICH) fields. More specifically, the preamble field appears as the first OFDM symbol in the frame and it is used for synchronization. The FCH field identifies the frame configuration information including the MAP message length and coding scheme and usable subchannels. DL-MAP and UL-MAP provide subchannel allocation for the DL and the UL subframes while the UL ranging subchannel, which is allocated for the mobile stations (MSs), is used for closed-loop time, frequency, and power adjustments as well as bandwidth requests. Finally, the UL CQICH enables the MSs feedbacking the channel-state information to the managing entities.

The WiMAX forum included advanced physical layer features to enhance the mobile WiMAX network coverage and capacity. For instance, mobile WiMAX supports the mandatory QPSK, 16 quadrature amplitude modulation (QAM), 64 QAM schemes in the DL, and the optional 64 QAM in the DL. Meanwhile, both convolutional code (CC) and convolutional turbo code (CTC) coding schemes are supported along with two other optional coding schemes, which are the block turbo code (BTC) and low density parity check code (LDPC). The combination of various modulation and code rates results in a fine resolution of data rates so that the BS scheduler may determine the best suited data rate for each burst allocation, based on the buffer size and the channel propagation conditions at the receiver. Moreover, a channel quality indicator (CQI) channel is used for providing channel-state information from the MSs to the base station scheduler while other channel-state information can be provided to the BS by the CQICH, which includes the physical carrier to interference plus noise ratio (CINR), the effective CINR, the MIMO mode selection, and the frequency selective subchannel selection. Hybrid auto repeat request (HARQ) is supported to provide fast response to packet errors and improve the cell edge coverage through using N channel Stop and Wait protocol. The previously stated adaptive modulation and coding, CQICH and HARQ, guarantee robust link adaptation in mobile environments at vehicular speeds exceeding 120 km/h.

THE OSI MODEL AND HANDOFF

A layer 2 handoff consists in a change of the layer 2 information related to the current attachment between the MN and the network. This information may consist in a layer 2 address or a layer 2 PoA. A trivial example of layer 2 handoff is performed when a laptop is associated with a WLAN AP and then switches to a second WLAN AP in the same subnet. Layer 2 handoff induces messages exchange between the affected MN and the previous and target PoAs. The time period of the first and the last layer 2 signaling message is referred to as layer 2 handoff delay, after which higher-layer protocols can proceed with their signaling procedure. This delay is a built-in value for a particular access technology.

Layer 3 handover consists in a change of layer 3 information related to the current attachment between the MN and the network. This information may consist in a layer 3 address such as an IP address, a routing table entry, or a point of service (PoS) such as an access router in WLANs. A trivial example is when a laptop is associated with an AP in one subnet and then switches to a second AP in a different subnet. When an MN changes its PoS during an active data session, all packets will continue to be routed to the old address unless particular mobility solutions are implemented. To address this issue, the IP packets may be reassociated with the same transport layer session; thus inducing important delays. IP tunneling may also be adopted. More specifically, a second address will be used as the tunnel end point while the old IP address will be maintained. IP tunneling, however, induces important overheads along with important delays. Inter-subnet mobility solutions addressing layer 3 handoff may be classified with respect to the protocols they use. One can distinguish IPv4-based, IPv6-based, or MPLS-based mobility solutions. Besides, it is worth noticing that some mobility solutions also called reactive solutions trigger the layer 3 handoff after the layer 2 handoff. Contrarily, predictive mobility solutions trigger the layer 3 handoff before the layer 2 handoff. Last but not least, some hierarchical mobility solutions define one global (interdomain) location update point and one local (intradomain) location update point to keep the intradomain handoff transparent to the global location update point and reduce the signaling overhead.

The Session Initiation Protocol (SIP) was designed to support mobility at the application layer; it is also used as the signaling protocol for real-time multimedia calls including voice over IP ones. As it acts at a higher layer, SIP can be combined with Mobile IP protocols, with the possibility of end-to-end adaptation for vertical handoff. A mobile host (MH) registers with a SIP server in its home domain. When a correspondent host (CH) sends an INVITE message to the MH to initiate a call, the redirect servers (RS), which have the current location of the MH, redirect the INVITE to the new location. When the MH moves during a session, it should send a new INVITE to the CH using the same call identifier in the original call setup, but it indicates the new IP address in the contact field of the SIP signaling message. The MH should also perform a new registration at the SIP server with its unique URI (uniform resource identifier) for all the new incoming calls. When both MH and CH are mobile, the reinvite messages are sent through the SIP server because it keeps track of the current CHs location.

HANDOFF VERSUS ROAMING

1. Handoff: Overview

The recent years have been marked by a growing need for advanced applications and Internet-related services at high throughput and low costs while guaranteeing continuous and open access to such services. Besides, mobile access to Internet services requires access to packet-switched wireless networks, which are generally divided into an access part and a core part. The access part contains the link-layer devices while the core part contains network-layer devices mainly acting at the network layer. A mobile node (MN) wishing to access Internet services needs to attach to the first link-layer device in the access network, also called, link-layer point of attachment (PoA). Nevertheless, the communicating MN may change the link-layer PoA during an active session due to particular reasons. This process is called handover or handoff. Handoff may be mobility driven when link conditions change due to mobility. In such case, the handoff occurs when the MN leaves the radio coverage of the actual PoA and becomes threatened by losing the network connectivity. On the other hand, handoff may be policy driven when the change in the PoA is argued by higher data rates, better services, lower costs, etc. After selecting the new PoA, a particular mobility mechanism implementing the handover process is executed. Such a mobility mechanism updates the mobile device and the network states to reflect the new PoA and redirects the ongoing data packets that were addressed to the old PoA during the handoff. If the MN has switched to a new link-layer PoA belonging to the same access network, a link-layer mobility mechanism will be executed. However, if the mobile device changes between link-layer PoAs belonging to different access networks, a network-layer mobility mechanism will be executed. When both PoAs (the old and the new) implement the same physical layer technology, the handoff is referred to as intratechnology handoff or horizontal handoff. Besides, when the new PoA implements a different physical layer technology compared to the old PoA, an intertechnology handoff, vertical handoff, or roaming takes place. In the latter case, the MN should be equipped with more than one network interface card (NIC). On the other hand, the handoff, which involves connecting to a new link and disconnecting from an old one, may be hard, soft, smooth, or seamless depending on the implemented mobility solution. Hard handoff takes place when the MN receives data packets from only one PoA during the handoff as the current link is disconnected first and then the connection to a new link is established. Controversially, soft handoff enables the MN to receive data packets by more than a PoA simultaneously as first a connection is established to a new link and then the old link is disconnected. Finally, a smooth handoff is a handoff where packet loss is minimized while a seamless handoff is a handoff transparent to the application. Handoff is always triggered by the occurrence of a certain event at the mobile device or in the network. If the handoff trigger occurs in the mobile device, the handoff is mobile initiated; otherwise, it is network initiated. As there is more than one candidate PoA to switch to, the decision to select the new PoA may be taken by the MN or the network; thus the handoff may be mobile controlled or network controlled, respectively. The handoff is mobile controlled and network assisted when the handoff decision is taken by the mobile based on information received from the network. Controversially, if the handoff decision is taken by the network upon the reception of information from the mobile such as the signal quality of neighboring PoAs, the handoff is network controlled and mobile assisted.

2. Roaming: Overview

Roaming appeared with the Global System for Mobile (GSM) technology. It may be defined as the set of mechanisms that allows extending the connectivity service to a location that is not covered by the home network (HN) where the service is registered but is covered by a visited network (VN). The VN always asks the HN for the authentication and authorization data associated with the visiting MN to allow or deny service access. Roaming also refers to changing the access network while on move. For instance, an MN which is GSM enabled and equipped with a wireless local area network (WLAN) NIC may roam between the two networks depending on its location without interrupting an active session. Roaming support is provided through the implementation of mobility management, authentication, authorization, and billing procedures. Different service providers managing different networks need to negotiate a roaming agreement to define the legal aspects related to service availability and billing. Often, the roaming process consists of three steps. First, the HN detects the MN as an unknown device which is not registered; therefore, the VN tries to identify the MN’s HN. If there is no roaming agreement between the HN and the VN, the MN will be not able to access VN services. However, if a roaming agreement exists, the VN requests MN’s service information to deduce whether the MN is allowed to roam. If it is the case, the VN maintains a subscriber record for the device while the HN updates the MN-related information so that any information destined to the MN will be correctly routed. The subscribers activity is registered in a file maintained by the VN. This file includes the details of the initiated calls, the subscribers visited locations, the calling parties, the volume of the exchanged data in case of data calls, the time of the calls and their duration, etc. Based on such information, the HN will be able to perform billing. Interstandards roaming allows MNs to seamlessly move between mobile networks with different access technologies. Nevertheless, this roaming type is particularly challenging as communication technologies have evolved independently across continents and were implemented by different industry bodies.