QoS-LTE-IEEE16m

1

QoS in LTE and 802.16Karthik R.M., Nadeem Akhtar,

CEWiTemail: [email protected], [email protected],

Jaydip Sen , Arijit Ukil,Innovation Lab, Tata Consultancy Services Ltd.

email: [email protected] , [email protected]

I. INTRODUCTION

Quality of Service (QoS) is a broad and loose term that refers to the “collective effect of service”, as perceived bythe user. For the purposes of this discussion, QoS more narrowly refers to meeting certain requirements typically,throughput, packet error rate, delay, and jitter-associated with a given application. Broadband wireless networksmust support a variety of applications, such as voice, data, video, and multimedia, and each of these has differenttraffic patterns and QoS requirements. In addition to the application-specific QoS requirements, networks oftenneed to also enforce policy-based QoS, such as giving differentiated services to users based on their subscribedservice plans. The variability in the QoS requirements across applications, services, and users makes it a challengeto accommodate all these on a single-access network, particularly wireless networks, where bandwidth is at apremium. From a user perspective, however, the perceived quality is based on the end-to-end performance of thenetwork. To be effective, therefore, QoS has to be delivered end-to-end across the network, which may include,besides the wireless link, a variety of aggregation, switching, and routing elements between the communication endpoints. IP-based networks are expected to form the bulk of the core network; hence, IP (Internet Protocol)-layerQoS is critical to providing end-to-end service quality.

IEEE Standard 802.16 [1] defines the air interface specification for wireless metropolitan area networks (WMANs).IEEE Standard 802.16 is designed to evolve as a set of interfaces based on a common Medium Access Control(MAC) protocol but with physical layer specifications dependent on the spectrum of use and associated regulations.The access and bandwidth must accommodate multiple end users. The services required by these end users are variedin their nature and include legacy time-division multiplex (TDM) voice and data, Internet Protocol (IP) connectivity,and packetized Voice-over-IP (VoIP). To support this variety of services, the 802.16 MAC must accommodate bothcontinuous and bursty traffic. Additionally, these services expect to be assigned QoS in keeping with the traffictypes.

A broad industry consortium, the Worldwide Interoperability for Microwave Access (WiMAX) Forum has beguncertifying broadband wireless products for interoperability and compliance with IEEE 802.16 standard. The WiMAXForum defines a limited number of system profiles and certification profiles. The system profile defines the subsetof mandatory and optional physical and MAC layer features selected by the WiMAX Forum from the IEEE 802.16standard. A certification profile is defined as a particular instantiation of a system profile, where the operatingfrequency, channel bandwidth, and duplexing mode are also specified.

Third generation Universal Mobile Telecommunications Systems (UMTS) based on Wideband Code DivisionMultiple Access (WCDMA) has been deployed widely. To ensure that this system remains competitive in thefuture, 3GPP (Third Generation Partnership Project) started a project to define the Long Term Evolution (LTE) ofUMTS cellular technology. The specifications related to this effort are known as evolved UMTS terrestrial radioaccess (E-UTRA) but are commonly referred to by the project name, LTE. Evolved Packet System (EPS) is thename given to the IP-based core network architecture defined in Release 8 of the 3GPP specifications. EPS isthe evolution from the General Packet Radio Service (GPRS)-based core network architecture used in UMTS/3Gnetworks. Compared to GPRS core network, EPS is much simpler in terms of the number of network elementsand flatter as well. In [2], integration between mobile WiMAX and 3GPP using the Policy and Charging Control(PCC) framework has been studied and a roaming architecture for WiMAX-3GPP integration is also proposed.

In order to provide end-to-end QoS, we need to go beyond the air-interface and look at broadband wireless systemsfrom an end-to-end network perspective. We need to look at the overall network architecture, higher-layer protocols,

and the interaction among several network elements beyond the mobile station and the base station. Providing end-to-end QoS requires mechanisms in both the control plane and the data plane. Control plane mechanisms are neededto allow the users and the network to negotiate and agree on the required QoS specifications, identify which usersand applications are entitled to what type of QoS, and let the network appropriately allocate resources to eachservice. Data plane mechanisms are required to enforce the agreed-on QoS requirements by controlling the amountof network resources that each application/user can consume.

In this document, we provide a brief overview of QoS related issues in LTE and WiMAX, ,identify somelimitations in the current standards and also propose some extensions required to overcome those limitations. Withrespect to LTE, we briefly discuss the bearers associated with LTE, QOS requirements in various applications, thebearer establishment procedures, and terminal and network-initiated QoS control. For WiMAX, we describe theQoS architecture and discuss a mapping mechanism between IP Differentiated Services (DiffServ) traffic classesand 802.16 service classes.

The rest of the document is organized as follows. Section II discusses QoS issues in LTE. It first presentsa brief introductory concept on QoS in LTE followed by issues like EPS bearers, GBR and non-GBR bearers,default and dedicated bearers. Various QoS requirements like user differentiation, fast session startup, and backwardcompatibility issues are discussed next. The details of a bearer establishment procedure is also presented. Finally,the section discusses procedures for network- and device-initiated QoS control in used in LTE. In Section III, webriefly discuss the IEEE 802.16 architecture and describe the QoS classes. Section III-B overviews the WiMAXframework for QoS, gives a brief description of the QoS functional elements and also highlights the requirements tobe met and extensions in the existing standard. In Section IV, we discuss the issues involved in ensuring end-to-endQoS and describes the mapping between 802.16 QoS classes and the DiffServ classes. Section V concludes thereport.

II. QOS IN LTE

A. Introduction

QoS in LTE provides access network operators and service operators with a set of tools to enable serviceand subscriber differentiation. LTE core network is known as evolved packet system (EPS) for its support toall-IP configuration. QoS in LTE is primarily network-initiated and class-based, where a service is offered to asubscriber by the operator. The term, “service” is used as the offering an operator makes to a subscriber. Basically,QoS mechanisms allow the access operator to enable service and subscriber differentiation, as depicted in Fig. 1.Examples of a service include VoIP telephony based on the IP Multimedia Subsystem (IMS), Mobile Television,Internet-Access (with various levels of user differentiation), Instant Messaging, Multimedia Broadcast MulticastService (MBMS) and Push-to-Talk over Cellular (PoC). We further need to distinguish between session-basedservices and non-session-based services. Session-based services utilize an end-to-end session control protocol suchas SIP/SDP or RTSP/SDP. All IMS services are session-based, while Internet-Access is an example of a non session-based service. The traffic running between a particular client application and a service can be differentiated intoseparate service data flows. For example, an IMS-VoIP session can be differentiated into two service data flows, onefor the session control signaling, and one for the media. The term, Traffic Forwarding Policy (TFP) denotes a setof pre-configured traffic handling attributes relevant within a particular user plane network element. For example, aRAN-TFP may include several attributes, such as the link layer protocol mode (acknowledged or unacknowledged),the power settings, and a default uplink maximum bit rate; while a Gateway TFP (GWTFP) may only include adefault downlink maximum bit rate. Each edge/bottleneck node potentially includes transport network node, whichsupports a number of TFPs. Uplink (UL) and Downlink (DL) Guaranteed Bit Rates (GBRs) are not part of a TFP,since these traffic handling attributes cannot be preconfigured for a QoS class. They must therefore, instead bedynamically signaled. TFPs confine traffic handling attributes to those nodes where those attributes are actuallyneeded. TFPs are provided and configurable by the operator from the management plane, as shown in Fig. 2.

LTE supports “end-to-end” QoS, meaning that bearer characteristics are defined and controlled throughout theduration of a session between the mobile device (UE) and the gateway (GW). QoS in LTE is characterized by anindex, QoS Class Identifier (QCI), and the parameter Allocation and Retention Priority (ARP). Bearer types belongto two main classes with guaranteed and non-guaranteed rates, which specify in more detail the values of packetdelay and loss that can be tolerated for any given bearer.

Fig. 1: Service and subscriber differentiation in LTE [3]

Fig. 2: Key elements for providing QoS in an access network [4]

B. EPS Bearer

EPS bearer uniquely identifies packet flows that receive a common QoS treatment between gateway and theterminal. A bearer is the level of granularity for QoS control in EPS based LTE. One bearer exists per combinationof QoS class and IP address of the terminal. The bearer is the basic enabler for traffic separation to providedifferential treatment for traffic with differing QoS requirements. As per functionality, two types of bearers exist:GBR and non-GBR and as per configuration, two another types of bearers exist: default and dedicated bearers.A bearer, in general is referred to as an edge-to-edge association between the UE and the GW. Independent ofwhether it is realized in a connection-oriented or a connectionless way, a bearer is defined through:

1) Network to which it connects the UE (referred to as Access Point Name in 3GPP),2) QoS Class Identifier (QCI) via which it can be associated with a TFP defined within each user plane

edge/bottleneck node, and3) (Optionally) the UL- and DL-GBR. Within an access network the UL-GBR and DL-GBR are only relevant

for session-based services, and only if the operators policy defined for a specific QoS class requires thatsession admission control (e.g., in the RAN) be triggered when establishing service data flows associated

with that QoS class.The term, QCI is not associated with any semantics, e.g., related to traffic characteristics or application layerrequirements on end-to-end QoS. That is, a QCI is simply a “pointer” to a TFP. Note further that, within a specificnode, multiple QCIs may be associated with the same TFP. In order to receive a QoS level other than the defaultQoS level (via the default bearer explained below), a service data flow needs to be bound which is referred to asa QoS bearer.

C. GBR and non-GBR bearers

For GBR bearers, maximum bit rate (MBR) and GBR are defined, though in 3GPP Release 8 [5], MBR and GBRare identical. GBR bearers are shielded from congestion related packet loss. This is realized by admission controlfunctions residing in different network nodes, like LTE base station and is executed at the point when a beareris established and modified. A GBR bearer is normally established “on demand”, because it blocks transmissionresources by reserving them in an admission control function, where as a non-GBR bearer can remain establishedfor long periods of time because it does not block transmission resources. GBR bearer is set up for premiumusers, where preference is blocking a new service request rather than risking degrading the performance of thealready admitted service request. It is an operator’s policy decision to establish GBR bearer for a service request,primarily based on expected traffic load versus predicted capacity. If sufficient capacity or low expected traffic loadis assumed, any service both real-time and non real-time, can be realized on non-GBR bearers.

D. Default and dedicated bearers

Default bearers are established when the user terminal is initially attached to the network. One default bearerexists per IP address to provide basic connectivity. As the default bearer can remain established for long periods, inLTE it is mandatory that default bearer be a non-GBR bearer. Dedicated bearers are mostly for GBR services, buteven non-GBR bearers can also be dedicated bearer depending on the policy undertaken by the operator based onits objective function and constraints. The operator maps the packet flows onto the dedicated bearer based on thepolicies provisioned into the network by Policy and Charging Resource Function (PCRF). It also evaluates the QoSlevel of the dedicated bearer. Conceptual depiction of a terminal with a default and dedicated bearer establishedonto the same IP address is shown in Fig. 3, where it is considered that end-to-end IP packet entering the system isprovided with a tunnel header. A tunnel header contains the bearer identifier to associate the packet with the correctQoS parameter, which is based on subscriber differentiation (business vs. standard, post-paid vs. pre-paid, roamers,privileged (civil administrator, police) and service differentiation (IP multimedia sub-system (IMS) voice, P2P filesharing, real-time audio, video streaming, mobile-TV). In the transport network, the tunnel header further containsdiffserv code point (DSCP) value (Fig. 3). In network layer, a packet flow is differentiated based on five-tuplepoints; namely, source and destination IP address, source and destination port numbers and protocol ID.

E. QoS Requirements in LTE

LTE QoS is evolved based on few strategic requirements to satisfy and achieve the targets like high data rate,low latency, and higher aggregate throughput. They are:

1) Operator Controlled Service and User Differentiation2) Minimize Terminal Involvement in QoS and Policy Control3) QoS Support for Access Agnostic Client Applications (UE-Based + Non-UE-Based)4) Fast Session Setup5) Backwards Compatibility1) Operator Controlled Service and User Differentiation: Service and user differentiation requires a limited

set of well defined QoS classes. The number of QoS classes supported within an operators network reflects thegranularity of differentiation the operator provides. Operators should be free to define the mapping of the servicedata flow(s) of offered services to the QCI(s). For certain well-known services this mapping could be standardized,or defined as part of roaming agreements. Likewise, operators should be free to define which TFP gets associatedwith a QCI.

Fig. 3: Bearer and associated QoS parameters in LTE

2) Minimize Terminal Involvement in QoS and Policy Control: Operators may regard a UE as a non-trusteddevice which can be “hacked”, e.g., for the purpose of receiving higher QoS than subscribed and charged for.Therefore, the control over a bearer’s QCI should be located within the network. In principle, there is no reason fora UE to have knowledge of a bearer’s QCI. Another aspect of this requirement is the placement of the exceptionhandling control associated with bearer establishment. To ensure a consistent exception handling across terminalsfrom different vendors, this control should be located within the network.

3) Support for Access Agnostic Client Applications (UE-Based + Non-UE-Based): Access agnostic client appli-cations do not use any vendor and/or access-specific QoS-API (Application Programming Interface). A QoS-APIcan be used to request the establishment of a QoS bearer, and thereby create the UL binding between a servicedata flow of the requesting client application and the QoS bearer. This requirement basically says that any clientapplication programmed towards the ubiquitous socket-API that is supported by virtually every widely deployedoperating system should be able to receive QoS. Note that the socket-API does not support requests for QoS bearers.

4) Fast Session Setup: It is widely recognized that low session setup delays are an important factor in userperceived service quality.

5) Backwards Compatibility: It can be expected that UEs based on the 3GPP LTE QoS concept will be widelydeployed in the coming years. Also, the upgrade of network equipment can not be assumed to be carried out “overnight”. Hence, backwards compatibility with LTE based equipment needs to be ensured by an evolved 3GPP QoSconcept.

6) QoS Class Identifier (QCI): QCI is a scalar that is used within the access networks as a reference to node-specific parameters that control packet-forwarding function, like resource allocation constraints, scheduling weights,queue management, buffer size) and each bearer is assigned one and only one QCI and uniquely identified by it.QCI characteristics are generally used to describe bearer type (GBR, non-GBR), priority, packet-delay budget, andpacket-error-loss-rate. In the access network, it is the responsibility of the eNodeB to ensure the necessary QoS fora bearer over the radio interface. Each bearer has an associated QCI, and an Allocation Retention Priority (ARP).Each QCI is characterized by priority, packet delay budget and acceptable packet loss rate. The QCI label for abearer determines how it is handled in the eNodeB. Only a dozen such QCIs have been standardized so that vendorscan all have the same understanding of the underlying service characteristics and thus provide the correspondingtreatment, including queue management, conditioning and policing strategy. This ensures that an LTE operator canexpect uniform traffic handling behavior throughout the network regardless of the manufacturers of the eNodeBequipment. The set of standardized QCIs and their characteristics (from which the PCRF in an EPS can select) isprovided in Table I. The QCI table specifies values for the priority handling, acceptable delay budget and packeterror loss rate for each QCI label.

The priority and packet delay budget (and to some extent the acceptable packet loss rate) from the QCI label

TABLE I: Standardized QoS Class Identifiers (QCIs)for LTE [5]QCI Resource Priority Packet delay (ms) Packet loss Services1 GBR 2 100 10−2 Conversational voice2 GBR 4 150 10−3 Conversational voice (live streaming)3 GBR 3 50 10−3 Real-time gaming4 GBR 5 300 10−6 Non-conversational video (buffered streaming)5 Non-GBR 1 100 10−3 IMS signaling6 Non-GBR 6 300 10−6 Video (buffered streaming)7 Non-GBR 7 100 10−3 Voice, video (live streaming), interactive streaming8 Non-GBR 8 300 10−6 TCP-based (e.g. WWW, e-mail), FTP, P2P, etc.,9 Non-GBR 9 300 10−6

determine the RLC mode configuration and how the scheduler in the MAC handles packets sent over the bearer(e.g., in terms of scheduling policy, queue management policy and rate shaping policy). For example, a packetwith a higher priority can be expected to be scheduled before a packet with lower priority. For bearers with alow acceptable loss rate, an Acknowledged Mode (AM) can be used within the RLC protocol layer to ensure thatpackets are delivered successfully across the radio interface.

The ARP of a bearer is used for call admission control- i.e., to decide whether or not the requested bearer shouldbe established in case of radio congestion. It also governs the prioritization of the bearer for pre-emption withrespect to a new bearer establishment request. Once successfully established, a bearer’s ARP does not have anyimpact on the bearer-level packet forwarding treatment should be solely determined by the other bearer level QoSparameters such as QCI, GBR and MBR.

An EPS bearer has two cross multiple interfaces as shown in Fig. 4 the S5/S8 interface from the P-GW to S-GW,the S1 interface from the S-GW to the eNodeB, and the radio interface (also known as the LTE-Uu interface) fromthe eNodeB to the UE. Across each interface, the EPS bearer is mapped onto a lower layer bearer, each with itsown bearer identity. Each node must keep track of the binding between the bearer IDs across its different interfaces.

An S5/S8 bearer transports the packets of a EPS bearer between a S-GW and an eNodeB. A radio bearertransports the packets of an EPS bearer between a UE and an eNodeB. An eNodeB stores a one-to-one mappingbetween a radio bearer ID and an S1 bearer to create the mapping between the two.

Fig. 4: LTE/SAE bearers across different interfaces [5]

IP packets mapped to the same EPS bearer receive the same bearer-level packet forwarding treatment (e.g.,scheduling policy, queue management policy, rate shaping policy, RLC configuration). Providing different bearer-level QoS thus requires that a separate EPS bearer is established for each QoS flow, and the user IP packets mustbe filtered into the different EPS bearers.

Packet filtering into different bearers is based on Traffic Flow Templates (TFTs). The TFTs use IP headerinformation such as source and destination IP addresses and Transmission Control Protocol (TCP) port numbersto filter packets such as VoIP from web browsing traffic so that each can be sent down the respective bearers withappropriate QoS. An uplink TFT (UL TFT) associated with each bearer in the UE filters IP packets to the EPSbearers in the uplink direction. A downlink TFT (DL TFT) in the P-GW is a similar set of downlink packet filters.

As a part of the procedure by which a UE attaches to the network, the UE is assigned an IP address by theP-GW and at least one bearer is established. This is called the default bearer, and it remains established throughoutthe lifetime of the PDN connection in order to provide the UE with always-on IP connectivity to that PDN. Theinitial bearer-level QoS parameter values of the default bearer are assigned by the MME, based on the subscriptiondata retrieved from the HSS. The PCEF may change these values in interaction with the PCRF or according tolocal configuration. Additional bearers called dedicated bearers can also be established at any time during or aftercompletion of the attach procedure. A dedicated bearer can be either a GBR or a non-GBR bearer, (the defaultbearer always has to be a non-GBR bearer since it is permanently established). The distinction between defaultand dedicated bearers should be transparent to the access network (e.g. E-UTRAN). Each bearer has an associatedQoS, and if more than one bearer is established for a given UE, then each bearer must also be associated withappropriate TFTs. These dedicated bearers could be established by the network, based for example, on a triggerfrom the IMS domain, or they could be requested by the UE. The dedicated bearer for a UE may be providedby one or more P-GWs. The bearer-level QoS parameter values for dedicated bearers are received by the P-GWfrom the PCRF and forwarded to the S-GW. The MME only transparently forwards those values received from theS-GW over the S11 reference point to the E-UTRAN.

F. Bearer Establishment Procedure

This section describes an example of the end-to-end bearer establishment procedure across the network nodesusing the functionality described in the previous sections.

A typical bearer establishment flow is shown in Fig. 5. Each of the messages is described below.When a bearer is established, the bearers across each of the interfaces discussed above are established. The PCRF

sends a ‘PCC (policy Control Changing) Decision Provision’ message indicating the required QoS for the bearerto the P-GW. The P-GW uses this QoS policy to assign the bearer-level QoS parameters. The P-GW then sendsa Create Dedicated Bearer Request message including the QoS and UL TFT to be used in the UE to the S-GW.The S-GW forwards the ‘Create Dedicated Bearer Request’ message (including QoS, UL TFT and S1-bearer ID)to the MME (message 3 in Fig. 5). The MME then builds a set of session management configuration informationincluding the UL TFT and EPS bearer identity, and includes it is the ’Bearer Setup Request’ message which itsends to the eNodeB (message 4 in Fig. 5). The session management configuration in NAS information and istherefore sent transparently by the eNodeB to the UE.

The Bearer Setup Request also provides the QoS of the bearer to the eNodeB; this information is used by theeNodeB for call admission control and also to ensure the necessary QoS by appropriate scheduling of the user’sIP packets. The eNodeB maps the EPs bearer QoS to the radio bearer QoS. It then signals a ‘RRC ConnectionReconfiguration’ message (including the radio bearer QoS, session management configuration and EPs radio beareridentity) to the UE to set up the radio bearer (message 5 in Fig. 5). The RRC Connection Reconfiguration messagecontains all the configuration parameters for the radio interface. This is mainly for the configuration of the layer 2(the PDPC, RLC and MAC parameters), but the Layer 1 parameters required for the UE to initialize the protocolstack. Message 6 to 10 are the corresponding messages to confirm that the bearers have been set up correctly.

G. Network and Terminal-Initiated QoS Control

In the earlier development stage of 3GPP LTE, only terminal-initiated QoS control mechanism is proposed andaccepted. The philosophy behind terminal-initiated QoS control is based on the assumption that the informationabout the requested service (e.g., application-layer QoS requirements) is only present in the UE, and that thisinformation must be provided to the network from the requesting client application. This is depicted in Fig. 6. Thishas led to a number of important consequences which are perceived as limitations:

1) QoS bearer can only be initiated from the UE.2) The uplink and downlink binding states are controlled from the UE.

Fig. 5: A message flow diagram for LTE/SAE bearer establishment [5]

3) The QCI is represented as a record of attributes.One limitation resulting from UE-initiated establishment of QoS bearers is that it leaves the exception handling

to the UE’s local policy. For example, this policy could be to retry setting up the QoS bearer with the same ordifferent Requested QoS, or perhaps to simply give up trying to set up the call. This is in conflict with requirement2. Furthermore, the absence of the possibility to initiate the establishment of a QoS bearer from the network (usinga bearer handling procedure) precludes the possibility to pre-activate QoS bearers based on operator policy. Thisis in conflict with requirement 4. This leads to the development of procedures for network-initiated QoS control.In network-initiated QoS control, the network initiates the signal to set up a dedicated bearer with a specific QoStoward the terminal or UE and the RAN, which generally consists of base stations and gateways. The signal istriggered by an application function (AF) and deep-packet inspection (DPI) function. Using this paradigm, theclient application is left with “QoS unaware”, which is shown in Fig. 7. The main motivation for specifying thenetwork-initiated QoS control is that most of the IP related services (Internet access, IP-TV, IMS voice) are typicallyprovided by the access network operator through some third-party peering agreement. The advantage of network-initiated QoS control is that it minimizes the terminal involvement in QoS and policy control which can be usedto provide QoS to access-agnostic client applications, such as applications that can be downloaded and installedby the subscriber without any terminal vendor specific QoS-API. Another advantage is that client applications canreside in a node (like laptop, set top box) that is physically separated from the actual terminal to implement theconcept of “split-terminal”.

In this section, we discussed about QoS in LTE, defined the various bearers, specified the QoS requirements inLTE, gave the standardized QCI values and also described the bearer-establishment procedure. We also highlightedthe limitations of UE-initiated establishment of QoS bearers and the advantages of network-initiated QoS control.In the next section, we discuss the QoS in 802.16 and describe the WiMAX QoS framework and identify extensions

Application

Service

info Initiate QoS bearer (Requested QoS+DL binding)

UE RANNetwork

Service

Initiate QoS bearer

(Negotiated QoS)

Fig. 6: Terminal-initiated QoS control [4]

Fig. 7: Terminal-initiated (top) and network-initiated (bottom) QoS control [3]

required in the existing standard.

III. QOS IN 802.16

A. QoS classes in 802.16

The basic IEEE 802.16 architecture consists of one Base Station (BS) and one (or more) Subscriber Station (SS).BS acts as a central entity to transfer all the data from SSs in a Point to multipoint (PMP) mode. Transmissions takeplace through two independent channels: Downlink Channel (from BS to SS) and Uplink Channel (from SS to BS).Uplink Channel is shared between all SSs while Downlink Channel is used only by BS. WiMAX Network WorkingGroup (NWG) Release 1.0.0 specification [6] supports the Time Division Duplexing (TDD) mode of operations.The IEEE 802.16 is connection oriented. Each packet has to be associated with a connection at MAC level. Thisprovides a way for bandwidth request, association of QoS and other traffic parameters and data transfer relatedactions.

Scheduling services represent the data handling mechanisms supported by the MAC scheduler for data transporton a connection. Each connection is associated with a single scheduling service. A scheduling service is determinedby a set of QoS parameters that quantify aspects of its behaviour. These parameters are managed using MAC dialogmessages. There are 5 types of scheduling service on the uplink namely, Unsolicited Grant Service (UGS), extendedreal-time polling service (ertPS), real-time polling service (rtPS), Non-real-time Polling service (nrtPs) and besteffort (BE) service. Each service is associated with a set of QoS parameters that quantify aspects of its behaviour.A detailed description of each of the services is given in [1].

The parameters associated with each of the services are given in Table II.

TABLE II: Parameters of Scheduling services and typical applicationsClass of Service Parameters ApplicationUGS Maximum Sustained Traffic Rate, Maximum Latency and Tolerated Jitter E1/T1ertPS Minimum Reserved Traffic Rate, Maximum Sustained Traffic Rate, Maximum Latency Silent suppressed VoIPrtPS Minimum Reserved Traffic Rate, Maximum Sustained Traffic Rate, Maximum Latency MPEG videonrtPS Minimum Reserved Traffic Rate, Maximum Sustained Traffic Rate, Maximum Latency FTPBE No minimum service level requirement e-mail

B. WiMAX Forum QoS Architecture

The WiMAX Forum’s NWG is responsible for developing the end-to-end network requirements, architecture, andprotocols for WiMAX. The WiMAX NWG has developed a network reference model (NRM) based on an IP servicemodel to serve as an architecture framework for WiMAX deployments for supporting fixed, nomadic, and mobiledeployments and to ensure interoperability among various WiMAX equipment and operators. The NRM definesa number of functional entities and interfaces between those entities. The interfaces are referred to as referencepoints. The overall network can be divided into three parts:

1) mobile stations used by the end user to access the network2) the access service network (ASN), which comprises one or more base stations and one or more ASN gateways

that form the radio access network at the edge, and3) the connectivity services network (CSN), which provides IP connectivity and the IP core network functions.The QoS architecture framework [6] extends the IEEE 802.16 QoS model by defining the various QoS-related

functional entities in the WiMAX network and the mechanisms for provisioning and managing the various serviceflows and their associated policies. The WiMAX QoS framework supports simultaneous use of a diverse set ofIP services, such as differentiated levels of QoS on a per user and per service flow basis, admission control, andbandwidth management and calls for the use of standard Internet Engineering Task Force (IETF) mechanisms formanaging policy decisions and policy enforcement between operators. The NWG Release 1.0.0 specification [6]defines the following procedures for QoS provisioning:

1) Pre-provisioned service flow creation, modification, and deletion.2) Initial Service Flow creation, modification and deletion.3) QoS policy provisioning between AAA(Authentication, Authorization and Accounting) and SFA (Service

Flow Authorization).The IEEE 802.16 specification [1] defines a QoS framework for the air interface. It consists of the followingelements:

• Connection-oriented service.• 5 data delivery services namely UGS, ertPS, rtPS, nrtPS, and BE.• Provisioned QoS parameters for each subscriber.• A policy requirement for admitting new service flow requests.

Under the IEEE 802.16 specification, a subscription can be associated with a number of service flows characterisedby QoS parameters. This information is provisioned in a subscriber management system, namely, AAA or policyserver. There are two types of service models, namely, static service model and dynamic service model. In thestatic service model, the subscriber station is not allowed to change the parameters of provisioned service flowsor create service flows dynamically. A dynamic service flow request is triggered by mechanisms not specified in

802.16 specification and they are evaluated to decide whether the service flow request can be authorized. IEEE802.16 specification defines the following procedure for dynamic service flow creation:

• Permitted service flows and their associated QoS parameters provisioned via the management plane.• Service flow request evaluated against the provisioned information and created if permissible.• Service flow created transitions to an admitted state and finally to an active state. Transition to admitted state-

invocation of admission control in BS and soft resource reservation. Transition to active state- actual resourceassignment for the service flow. item Service flow can also transition in the reverse direction.

• A dynamically created service flow MAY be modified or deleted.1) QoS Functional elements: Based on the IEEE 802.16 [1] and the WiMAX reference model [6], the QoS

functional elements and their functions can be enumerated as follows:1) MS (Mobile Station) and ASN (Access Service Network). WiMAX network may support ASN-initiated

service flow creation.2) The home Policy Function (PF) and its associated policy rules belong into the H-NSP (Home Network Service

Provider). Maintained information include general policy rules and application specific policy rules. AAAprovisions the PF’s database with user’s QoS profile and associated policies.

3) AAA holds the users’ QoS profile and associated policy rules. They may be downloaded into the SFA (ServiceFlow Authorization) at network entry or into the PF.

4) Service Flow Management (SFM): is a logical entity in the ASN for creation, admission, activation, modifica-tion and deletion of 802.16 service flows. It consists of Admission Control (AC) and associated local resourceinformation. SFM always located in the BS. Admission Control is the ability to admit or ability to controladmission of a user to a network based on users service profile and network performance parameters (forexample, load and average delay). If a user requests access to network services but the incremental resourcesrequired to provide the grade of service specified in the users service profile are not available, the AdmissionControl function rejects the users access request. Admission Control is implemented to ensure service qualityand is different from authentication and authorization, which are also used to admit or deny network access.

5) SFA: There are two SFA-s, namely Anchor SFA and Serving SFA. Anchor SFA is assigned to each MS for theduration of the Device Authentication Procedure. Serving SFA directly communicates with the SFM. Both theSFAs know the identities of each other. SFA MAY perform ASN-level policy enforcement using local policydatabase and associated local policy function and admission control. Resource reservation request messagesare sent from the anchor SFA to the serving SFA and finally to the SFM. The result of the reservation is sentfrom the SFM to the serving SFA and finally back to the anchor SFA.

6) A network management system for administratively provisioning service flows.7) Pre-provisioned service flow: Set of service flows can be created, admitted and activated after an SS registers

with the WiMAX network.The MS initiates a call setup procedure. The QoS profile and policies are stored in the PF or SFA by the AAA.They evaluate the Service Flow Trigger requests and send the result of the evaluation back to the triggering elementwhich can be MS or the ASN itself.

The QoS functional elements [6] together with the procedure to setup a call are shown in Fig. 8.2) Extensions to the standard: The limitations of the existing Release 1 Version 1.2 [6] standard and the

extensions required can be enumerated as follows:1) The scope of the QoS in [6] is limited to the WiMAX radio link connection. QoS specific treatment in the

fixed part of core and access networks not specified. There are many possibilities for enforcing QoS in theaccess and core networks, and operators may require specific interfaces in ASN elements to use for mappingIP traffic onto these networks. Therefore, the QoS section in [6] makes no guarantees on end-to-end QoS.Operators require specific L2 and L3 interfaces in ASN network elements for mapping IP traffic on to thesenetworks.

2) In the specification [6], only preprovisioned service flows are defined. There is no scope for dynamic serviceflow creation. Dynamic service flows are triggered by the MS or AF. The AF issues service flow triggers tothe PF. The provision, admission and activation of dynamic service flows require interaction between PF andSFA. Hence, PF-SFA interaction not specified.

3) For a given ASN/NAP (Network Access Provider) there exists an anchor SFA assigned to each MS. The

MS

SFM

BS

AF

PF

AAA

Home or Visited NSP

Local resource

ASN

Service Flow Trigger Evaluation

done using policies in PF or SFA

Service request

ASN−GW

Admit or reject

Serving SFA

Anchor SFA

AdmissionControl

Result of

evaluation

Service

Flow Trigger

Admit

or reject

Service

Request

Resultof

evaluation

Service

Flow

trigger

Store QoS profile and

policies in PF or SFA

External

networks

Fig. 8: QoS functional elements and Call Setup

anchor SFA does not change for the duration of the Device Authentication session. Optionally, there may beone or more additional SFA entities that relay QoS related primitives and apply QoS policy for that MS. Therelay SFA that directly communicates with the SFM is called the serving SFA. Both the anchor and servingSFA know the identities of each other. The anchor and/or serving SFA may also perform ASN-level policyenforcement using a local policy database and an associated local policy function (LPF). The LPF can alsobe used to enforce admission control based on available resources. A serving SFA MAY be in the bearer pathtowards the SS, but only the signalling interactions for SFA are in the scope of [6]. Data path interactionsbetween PF and SFA are not defined.

4) Handoff capability from 3GPP to WiMAX is usually referred to as scenario 4 or inter-system handover.Seamless inter-system handover or scenario 5 provides greater service continuity than that perceived in intra3GPP handovers. However, both inter-system handover and seamless-inter-system handover are not addressedin WiMAX Release 1.0.

5) Maintaining a specific level of QoS consistently across the WiMAX and 3GPP access technologies involvesseveral considerations such as QoS mappings and semantics on the two access networks as well as appropriateresource allocations. WiMAX provides a powerful and flexible QoS handling using the QoS mechanisms from802.16 which is transparent to Direct IP access, However, QoS-enabled IP-based access networks cannot befully utilized within WiMAX-3GPP IP access.

3) Requirements: [1], [6], [7] have specified the general requirements of the Network Systems Architecture aswell as specific requirements from the QoS architectural framework. A summary of the general requirements is asfollows:

1) Architecture should support simultaneous set of diverse IP services including DiffServ and Integrated Services(IntServ), admission control and bandwidth management.

2) Policy enforcement per user based on the Service Level Agreements (SLAs) and also synchronisation betweenoperators based on SLA-s accommodating for the fact that not all operators implement the same policies.SLA-based resource management for subscribers should also be supported.

3) The architecture should be capable of supporting voice, multimedia services and other mandated regulatoryservices such as emergency services and lawful interception and should be agnostic to a variety of independentApplication Service Provider (ASP) networks.

4) Architecture should support interworking with existing wireless network using protocols based on IETF and

IEEE suite of protocols.5) The architecture does not preclude inter-technology handovers- e.g., to Wi-Fi, 3GPP- when such capability

is enabled in multi-mode MS6) It should support roaming between NSP-s. The architecture should allow a single NAP to serve multiple

MSs using different private and public IP domains owned by different NSPs (except where solutions becometechnically infeasible). The NSP MAY be one operator or a group of operators. Seamless handover betweendifferent vehicular speeds have to be addressed.

7) Interfacing with various interworking and media gateways for delivering services over IP to WiMAX accessnetworks should be supported.

8) Global roaming across WiMAX operators with credential reuse, use of AAA for accounting and charging,and consolidated/common billing and settlement.

9) Specifications should specify the rules in situations in which pre-provisioned service flows cannot be createdor activated in the ASN. QoS framework should allow the communication of an attempt to pre-provision aservice flow from the ASN to the CSN. The procedure is dependent on the policies within the ASN and theagreement between NAP and NSP.

IV. END-TO-END QOS

In this section, we deal with application and connection level QoS and briefly describe the issues in providingend-to-end QoS. We also briefly describe the DiffServ mechanism for providing QoS in WiMAX networks.

A. Application level QoS and connection level QoS

In general, QoS in wireless networks is considered at two levels, i.e., at the application level and connectionlevel. Application level QoS is related to the perceived quality at the user end. A set of parameters, such asdelay/delay jitter, error/loss and throughput, etc., are used to describe application level QoS. Efficient packet accessand packet scheduling schemes play key roles in solving these QoS problems. Connection-level QoS is related toconnection establishment and management. It measures the connectivity and continuity of service in a wirelessnetwork, mostly by two parameters: the new-call-blocking probability, which measures service connectivity, andthe handoff-dropping probability, which measures service continuity during handoff. For a mobile user, dropping anongoing call is generally more unacceptable than blocking a new call request. Therefore, minimizing the handoff-dropping probability is usually a main objective in the wireless system design. On the other hand, the goal of anetwork service provider is to maximize the revenue by improving network resource utilization, which is usuallyassociated with minimizing the new-call-blocking probability while keeping the handoff dropping below a certainthreshold.

Connection level QoS is influenced by call admission control (CAC). In cellular wireless networks the utilizationof system resources by new calls is often kept below a threshold level to accommodate handoff connections becauseservice providers are obligated to provide a minimum QoS to subscribers even as they aim to maximize bandwidthutilization. Thus, when a mobile SS engaged in a call is handed off to a new cell, it may receive a higher priority forchannel allocation by the new BS than new calls originating in the cell. Even with resource reservation, connectionscan still be dropped due to fluctuations in the received SINRs at the mobile SSs, especially for those located nearthe edges of cells. To handle a multiservice WiMAX access network, it is very important to employ the CACmechanism. First, CAC is a crucial step for the provision of QoS guaranteed service, because it can prevent thesystem capacity from being overused. Second, CAC can help WiMAX access network to provide different types oftraffic load with different priorities by manipulating their blocking probabilities. To carry out the admission controlfor each subscribers local network, a CAC manager can be placed in a WiMAX base station. The CAC managerknows the uplink/downlink bandwidth capacity of any subscriber k from other modules in the same base station.When an application in subscriber k’s local network initiates a connection to the Internet, it sends connection requestto the CAC manager with upstream bandwidth requirement bU and downstream bandwidth requirement bD. Thenthe CAC manager commits admission control check on both uplink and downlink. In this respect, the call admissioncontrol in WiMAX access network is a two-dimensional CAC problem. The two-dimensional CAC problem canbe decomposed into two independent one-dimensional problems namely uplink CAC policy and downlink CACpolicy to make admission tests on both the uplink and downlink separately, and only the connection request passing

both admission tests is admitted finally. In general, uplink traffic is only a fraction of the downlink traffic inmost applications. Hence, downlink admission control plays a more important rule than uplink admission control.Depending on whether minimizing new-call-blocking probability or minimizing handoff-dropping probability is theobjective, we can design CAC algorithms to achieve the desired objective.

B. Issues in providing end-to-end QoS

The links between intermediate nodes of an end-to-end call may use a variety of layer 2 technologies, suchas ATM, frame relay, and Ethernet, each of which may have its own methods to provide QoS. Since WiMAX isenvisioned to provide end-to-end IP services and will likely be deployed using an IP core network, IP QoS and itsinteraction with the wireless link layer are what is most relevant to WiMAX network performance.

Resource limitations in the network is what makes providing assurances a challenge. Although typically, themost-constrained resource is the wireless link, the other intermediate nodes and links that have to be traversed foran end-to-end service also have resource limitations. 1 Each link has its own bandwidth-capacity limits, and eachnode has limited memory for Over-building the network to provide higher bandwidth capacity and larger buffersis an expensive and inefficient way to provide quality, particularly when the quality requirements are very high.Therefore, more clever methods for providing QoS must be devised and these methods must take into accountthe particular needs of the application or service and optimize the resources used. Different applications require adifferent mix of resources. For example, latency-intolerant applications require faster access to bandwidth resourcesand not memory, whereas latency-tolerant applications can use memory resources to avoid packets being dropped,while waiting for access to bandwidth resources. This fact may be exploited to deliver QoS efficiently. In short, aQoS-enabled network should provide guarantees appropriate for various application and service types while makingefficient use of network resources.

Ensuring end-to-end QoS requires mechanisms in both the control plane and the data plane [7]. Control planemechanisms are required to allow the users and the network to negotiate and agree on the required QoS, identifywhich users and applications are entitled to what type of QoS, and let the network allocate the resources accordingly,to each service. Data plane mechanisms are required to enforce the agreed-on QoS requirements by controlling theamount of network resources that each application/user can consume.

The WiMAX network bearer consists of a wireless bearer and an IP transmission bearer. The former providesa wireless access service by the IEEE 802.16 mechanism, and the IP bearer deploys DiffServ and MPLS (Multi-Protocol Label Switching) and so on, to guarantee QoS. Traditional IP networks were designed for best-effort dataand did not include any provision for QoS. Some form of QoS can be provided by relying on different end-to-endtransport protocols that run over IP. For example, TCP ensures that data packets are delivered end-to-end reliably.For ensuring end-to-end latency and throughput, QoS mechanisms need to be in place in the network layer, andtraditional IP did not have any. One mechanism for ensuring IP QoS is via Diff Serv (Differentiated Services)architecture.

C. DiffServ and 802.16 service mapping

Diffserv is an IP layer QoS mechanism, whereby IP packets are marked with diffserv code points at the networkpoint of entry and network elements enforce relative priority of packets based on their code points. The diffservmethodology allows network resources to be reserved for classes of traffic, rather than for individual flows. DiffServdefines a number of service classes and QoS mechanisms that are applied to packets in those service classes (calledPer Hop behaviour or PHB). The DiffServ Code Point (DSCP) is located in the IP packet header and is used todetermine the PHB. The standard PHB-s each have a unique DSCP associated with them. The DSCP is used todetermine the respective DiffServ behaviour the packet is to receive. Different types of applications have differenttraffic characteristics and require different QoS behaviours to be applied to them. The different DiffServ classesare Expedited Forwarding (EF) Class, Assured Forwarding (AF) Class, Class Selector (CS) and Default DiffServClass.

Mapping between between DiffServ and UMTS network should take place when the IP network and UMTSnetworks are interconnected. A similar method for mapping between DiffServ and WiMAX should also be devel-oped. Mapping to DiffServ classes: UGS class requires a fixed amount of service. ertPS and rtPS can be mapped

to Expedited Forwarding (EF) DiffServ. nrtPS can be mapped to Assured Forwarding (AF) while best effort canbe either mapped to the AF or default DiffServ behaviour.

In the context of the WiMAX air link, IP diffserv mechanism can be used to enforce priorities for packets withina service flow, or to establish service flows based on diffserv classes for a given subscriber. As an example, a singlepre-provisioned service flow for a subscriber can be used to carry multiple types of traffic, with relative precedenceestablished based on diffserv code points. On the other hand, service flows MAY be established dynamically tocarry different diffserv traffic classes. An example of this is the establishment of a UGS service flow dynamicallyto carry a voice call, where the voice traffic is marked with diffserv EF class. In the first case above, the diffservcode points are used to prioritize and schedule packet transmission within a service flow. The manner in whichthis is done is a matter of local implementation in the BS and the SS, subject to the prioritization rules of diffserv.In the second case, the diffserv code point is used to classify packets onto separate service flows. This scenariooccurs when packets entering the BS or the MS are already marked with diffserv code points by an application orsome prior network entity.

ertPS and rtPS services can be mapped to the EF DiffServ Class. nrtPS can be mapped to the AF based on theemission priority, discard priorities and their minimum requirements. BE can be mapped to either the AF classor the Default DiffServ Class. A similar procedure is used in [8] to perform the mapping between UMTS serviceclasses and DiffServ traffic classes.

V. CONCLUSION

We provided a brief description of QoS in LTE and the 802.16 traffic classes and the extension of IEEE 802.16QoS framework to WiMAX QoS architecture. In LTE the EPS provides UEs with IP connectivity to the packet datanetwork. The EPS supports multiple data flows with different QoS per UE for applications that need guaranteed delayand bit rate such as VoIP as well as best effort web browsing. The EPS network architecture, EPS bearers, togetherwith their associated QoS attributes provide a powerful framework for the provision of a variety of simultaneousservices to the end user. From the perspective of the network operator, the LTE systems is also breaking new groundin terms of its degree of support for self-optimization and self-organization of the network via the X2, S1 and Uuinterfaces, to facilitate deployment. In LTE, each logical channel has a corresponding QoS description which shouldinfluence the behavior of the eNodeB resource scheduling algorithm. Based on the evolution of the radio and trafficconditions, this QoS description could potentially be updated for each service in a long-term fashion. It is likelythat the mapping between the QoS descriptions of different services and the resource scheduling algorithm in theeNodeB will be a key differentiating factor between radio network equipment manufactures. In a heterogeneousnetworking environment, guaranteeing end-to-end QoS will invite special challenges. It includes among other issues,mapping of the QoS attributes of the access and core networks to the QoS class identifier values of the applications,and design of suitable inter-working and inter-operating (I&I) elements in the gateways

WiMAX NWG Release 1.0.0 specification supports only a static QoS model based on the concept of preprovi-sioned service model. Extensions are required in the Release 1.0.0 to support dynamic creation, modification anddeletion of service flows. New system profile features are needed to enable advanced services such as location-basedservices and multicast-broadcast services. In addition, MAC layer efficiency has to be improved by reducing theMAC layer overhead. The next generation of mobile WiMAX is expected to provide flexible deployment solutionssuch as multi-hop relay, femtocell, and multicarrier support as well as optimized coexistence and interworking withother access technologies such as WiFi and 3G systems. In addition, flexible spectrum deployment is desirable.MAC layer efficiency has to be improved by lowering the MAC overhead especially for applications such as VoIPtraffic.

There are many issues with ensuring end-to-end QoS. QoS needs to be considered at different levels, namely theapplication level and connection level. Mechanisms are required to ensure that the QoS is met at the different levels.In addition, the links between intermediate nodes of an end-to-end call may use a variety of layer 2 technologiesand proper mapping mechanisms between the QoS of the different layer needs to be developed. Ensuring end-to-endQoS also requires mechanisms in both the control and user plane. One method of ensuring IP QoS is via Diff-Serv.

REFERENCES

[1] IEEE 802.16: IEEE Standard for Air Interface for Broadband Wireless Systems, June 2008.

[2] P. Taaghol, A. K. Salkintzis, and J. Iyer, “Seamless integration of mobile WiMAX in 3GPP networks,” IEEE Communications Magazine,vol. 46, no. 10, pp. 74–85, October 2008.

[3] H. Ekstrom, “QoS control in the 3GPP evolved packet systems,” IEEE Communications Magazine, vol. 47, no. 2, pp. 76–83, February2009.

[4] R. Ludwig, H. Ekstrom, P. Willars, and N. Lundian, “An evolved 3GPP QoS concept,” in IEEE Vehicular Technology Conference, vol. 1,May 2006, pp. 388–392.

[5] 3GPP TS 23.203 V8.3.1, Technical Specification, Policy and charging control architecture (Release 8).[6] WiMAX Forum Network Architecture, Stage 2: Architecture Tenets, Reference Models and Reference Points, January 2008.[7] J. G. Andrews, A. Ghosh, and R. Muhamed, Fundamentals of WiMAX, Understanding Broadband Wireless Networking. Prentice Hall,

2007.[8] S.Maniatis, E. Nikolouzou1, I. Venieris, and E. Dimopoulos, “Diffserv-based traffic handling mechanisms for the umts core network,”

in IST Mobile and Wireless Telecommunications Summit, June 2002.