also to manage network resources. In other words ... gaming, streaming, and social networking ser- vices, have diverse ... specific information in application layer signaling. [2]. On the ... In the UL direction, all SF types except UGS involve some ...
ALASTI LAYOUT
4/22/10
1:57 PM
Page 104
TOPICS IN WIRELESS COMMUNICATIONS
Quality of Service in WiMAX and LTE Networks Mehdi Alasti and Behnam Neekzad, Clearwire Jie Hui and Rath Vannithamby, Intel Labs
ABSTRACT
104
A quality of service framework is a fundamental component of a 4G broadband wireless network for satisfactory service delivery of evolving Internet applications to end users, and managing the network resources. Today’s popular mobile Internet applications, such as voice, gaming, streaming, and social networking services, have diverse traffic characteristics and, consequently, different QoS requirements. A rather flexible QoS framework is highly desirable to be future-proof to deliver the incumbent as well as emerging mobile Internet applications. This article highlights QoS frameworks and features of OFDMA-based 4G technologies — IEEE 802.16e, IEEE 802.16m, and LTE — to support various applications’ QoS requirements. A few advanced QoS features such as new scheduling service (i.e., aGP), quick access, delayed bandwidth request, and priority controlled access in IEEE 802.16m are explained in detail. A brief comparison of the QoS framework of the aforementioned technologies is also provided.
satisfactory service delivery to end users and also to manage network resources. In other words, today’s popular Internet applications, including real-time and non-real-time traffic such as multimedia services and online gaming, have very different traffic patterns and distinct QoS requirements. The traffic patterns of these emerging Internet applications show non-periodic variable-sized packet arrivals. The traditional QoS framework is no longer efficient and/or sufficient to support these new mobile Internet applications with good or required user experience. The organization of the article is as follows. The next section reviews the key elements of the QoS framework in IEEE 802.16e. We then highlight some advanced features in IEEE 802.16m to improve performance of a WiMAX network compared to a legacy network based on IEEE 802.16e. We then explain QoS framework of the LTE wireless technology. We then provide a high-level comparison between QoS frameworks of these three 4G wireless technologies focusing on the air interface. The final section draws some conclusions.
INTRODUCTION
QOS IN IEEE 802.16E
As the number of mobile broadband subscribers and the traffic volume per subscriber are rapidly increasing, quality of service (QoS) is becoming significant as operators move from single to multiservice offerings, and emerging rich devices capable of running multimedia and gaming applications. Fourth-generation (4G) broadband wireless technologies such as IEEE 802.16e, IEEE 802.16m, and Third Generation Partnership Project (3GPP) Long Term Evolution (LTE) have been designed with different QoS frameworks and means to enable delivery of the evolving Internet applications. As the Internet evolves, Internet applications and associated traffic patterns are also evolving over time. Web 2.0 supports rich media applications such as interactive voice and video services, web audio/video streaming services, and online gaming services, with smart optimization engines at both the client and server sides [1]. QoS specifically for evolving Internet applications is a fundamental requirement to provide
The QoS framework in IEEE 802.16e is based on service flows (SFs). An SF is a logical unidirectional flow of packets between the access service network gateway (ASN-GW) and a mobile station (MS) with a particular set of QoS attributes (e.g., packet latency/jitter and throughput) identified by a connection ID [2]. Based on IEEE 802.16e, packets traversing the medium access control (MAC) interface are associated with SFs according to classifier rules. Figure 1 demonstrates SFs in IEEE 802.16e. Traffic mapping to appropriate SFs is done at the ASN-GW for downlink (DL) and at the MS for uplink (UL) directions, respectively. Between the ASN-GW and the base station (BS), the QoS of the SFs is supported by backhaul transport QoS. On the air interface, a BS scheduler provides QoS for DL, and cooperation between the BS and MS schedulers provides QoS for UL. This air interface scheduler at the MAC sublayer determines how radio resources are assigned among multiple SFs
0163-6804/10/$25.00 © 2010 IEEE
IEEE Communications Magazine • May 2010
ALASTI LAYOUT
4/22/10
1:57 PM
Page 105
Service 2
APP 2
APP 1
Air interface scheduler
Classifier
Service 3
Service 1
APP 3
Classifier
Backhaul transport BS
MS
Air interface
ASN-GW
Backhaul transport
Figure 1. Service flows in the WiMAX QoS framework.
based on QoS attributes. Resources assigned to an MS enable it to receive traffic over DL and transmit data over UL. Details of air interface scheduler operation are not specified by the standard; therefore, it is vendor-specific. Traffic classification and mapping from application packets onto SFs in WiMAX is done at the convergence sublayer (CS), based on protocol-specific packet matching criteria like a combination of five-tuple, such as source and destination IP addresses, source and destination port address, protocol, and differentiated services codepoint (DSCP) [2]. IEEE 802.16e supports both QoS control paradigms: network-initiated, where SF creation is initiated by the BS, and terminal-initiated, where SF creation is initiated by the MS. With network-initiated, an application function (AF) inside the network can trigger messaging signals to set up SFs with appropriate QoS attributes; consequently, the client application can be left access-agnostic, and there is no need for accessspecific information in application layer signaling [2]. On the other hand, with terminal-initiated QoS control, the MS requests creation of SFs with appropriate QoS attributes; hence, the client application is aware of the specifications of the access QoS model [2]. Network-initiated SF creation is a mandatory, but terminal-initiated SF creation is an optional capability of IEEE
IEEE Communications Magazine • May 2010
802.16e [2]. SFs may be created, changed, or deleted through a series of MAC management messages referred to as DSX (i.e., DSA, DSC, and DSD).
SERVICE FLOW TYPES IN IEEE 802.16E AND ASSOCIATED PARAMETERS IEEE 802.16e supports five SF types [2]: • Unsolicited grant service (UGS): Supports real-time traffic with fixed-size data packets on a periodic basis • Real-time polling service (rtPS): Supports real-time traffic with variable-size data packets on a periodic basis • Extended rtPS (ertPS): Supports real-time traffic that generates variable-size data packets on a periodic basis with a sequence of active and silence intervals • Non-real-time polling service (nrtPS): Supports delay-tolerant traffic that requires a minimum reserved rate • Best effort (BE) service: Supports regular data services The following and Table 1 summarize some key SF QoS attributes in the IEEE 802.16e standard and provide some targeted traffic types for each SF: • Maximum sustained traffic rate (MSTR): Defines capping rate level of an SF
105
ALASTI LAYOUT
4/22/10
Service flow type
1:57 PM
MRTR
UGS
Page 106
MSTR
Max latency
Max jitter
X
X
X X
ertPS
X
X
X
rtPS
X
X
X
nrtPS
X
BE
Traffic priority
Targeted traffic Constant bit rate (CBR) services, TDM services
X
VoIP with silence suppression or activity detection
X
Streaming audio and video
X
X
File transfers
X
X
Web browsing, email
Table 1. IEEE 802.16e service flow types, some key QoS parameters, and targeted traffic types.
• Maximum traffic burst: Defines the maximum continuous burst a system should accommodate for a service • Minimum reserved traffic rate (MRTR): Specifies the minimum rate guaranteed to an SF • Maximum latency: Specifies maximum packet delay over the air interface • Tolerated jitter: Specifies maximum packet delay variation (jitter) for an SF • Traffic priority: Can be exploited to adjust the priority of packets of different SFs based on a combination of subscribers’ profiles and services mapped to SFs • Unsolicited grant interval (UGI): Defines the time interval between successive data grant opportunities for an SF over DL • Unsolicited polling interval (UPI): Defines the maximal interval between successive polling grant opportunities for an SF over UL WiMAX uses a BE SF, referred as the initial SF (ISF), to establish IP connectivity during network entry before any packet transmission and reception.
AIR INTERFACE SCHEDULER The SF framework provides QoS granularity and inter-SF isolation over the air interface. The air interface scheduler is responsible for enforcing QoS by assigning DL and UL physical (PHY) layer resource blocks among SFs. This mechanism is called bandwidth allocation. A scheduling decision is determined based on appropriate SFs’ QoS state variables, like buffer lengths, elapsed packet delay, SFs’ QoS requirements such as MRTR and maximum latency, and radio frequency (RF) conditions of different MSs. In general: • SFs with shorter maximum latency or SFs with higher MRTR receive higher priorities in the scheduling decision. • SFs with late packets or long buffer lengths also,receive higher priorities in the scheduling decision. • MSs with better RF conditions receive higher priorities by the scheduler in order to improve overall sector throughput. However, an operator can adjust fairness to ensure MSs in poor RF conditions receive reasonable QoS. The air interface scheduler may differentiate between traffic flows within an SF by packet priority levels such as DSCP values
106
(intra-SF). Also, it may further utilize the traffic priority attribute of SFs to differentiate between traffic associated with SFs of the same type (inter-SF).
IEEE 802.16E BANDWIDTH REQUEST AND GRANT MECHANISM In the UL direction, all SF types except UGS involve some form of bandwidth request/grant mechanism for bandwidth allocation. In the DL, the BS scheduler has all the information about DL SF status for making the best scheduling decision. However, UL SF status information is distributed in MSs. Additionally, an MS may need to be assigned some small bandwidth to send UL SF status information to the BS. Therefore, some mechanisms (i.e., bandwidth request [BR]) are required to inform the BS scheduler of UL SF status (i.e., BR message). Basically, BR refers to a mechanism MSs use to indicate to the BS their bandwidth needs. In IEEE 802.16e there are a number of methods for sending a BR message to BS: • UL bandwidth requests: An MS indicates to a BS that one of its UL SFs need bandwidth that may be carried through a bandwidth request header or a grant management subheader for a piggyback request, or a BR indicator on the channel quality indication channel (CQICH) feedback. UL BR messages are per SF in order to inform the BS about MS’ traffic composition. There are two types of BR messages [2]: –Incremental: The BS adds the new request to the current perception of bandwidth needs of an SF. –Aggregate: The BS replaces the perception of bandwidth it needs for an SF with a new request. In order to avoid error accumulation in the BR mechanism, MS periodically uses aggregate BR message. Thus, BR mechanism in IEEE 802.16e is self-correcting and no acknowledgment is required for the BR message. There are several ways of requesting bandwidth: –In contention-based (random access) BR, there is no reserved dedicated resource for an MS to transmit data, and the MS uses a code-division multiple access (CDMA)based mechanism; so the BS allocates
IEEE Communications Magazine • May 2010
ALASTI LAYOUT
4/22/10
1:57 PM
Page 107
enough bandwidth for the BR message before sending the BR header. –In contention-free BR, a signaling header or a piggybacked bandwidth request is used by appending it to UL data transmission when there is available UL resource or when BS polls MSs [2]. In this case the MS is allocated sufficient bandwidth to send a BR message. Although the UPI is an SF parameter, the polling is always on a per MS basis. An MS with an active UGS connection is not polled, but the MS uses the poll me (PM) bit in the header of a UGS SF packet instead. Once the BS detects this request, it individually polls the MS to satisfy its request. The piggyback request mechanism allows an MS to perform incremental BR per SF using the piggyback field in the grant management subheader (GMSH) [2]. This mechanism avoids the BS allocating for the BR header. However, the capability of piggyback request is optional. • UL bandwidth grants: After the BS is informed of a UL SF status, it makes a scheduling decision and allocates some bandwidth to the SF. However, by this time the status of the UL SF might have changed due to new packet arrivals. Therefore, the bandwidth grant mechanism in IEEE 802.16e is on a per MS basis. In other words, the BS assigns a UL burst to the MS for all of its SFs. This allows real-time reaction of the MS to QoS needs for any redistribution of bandwidth among the SFs. Also, this simplifies the system by sending only one grant (UL MAP IE) per MS instead of a number of grants per each SF. An intelligent MS scheduler distributes the allocated bandwidth among its SFs. A voice service can be served using a UGS type SF with no need for BR during a talk spurt period; it does not need any BR during a silence period, either. However, there is a need for BR for a silence to talk spurt. IEEE 802.16e has provisioned ertPS for this purpose. Request/Grant for ertPS — The ertPS SF type is designed to reduce the complexity of the BR mechanism for some services such as VoIP with silence suppression. With ertPS and during a talk spurt, a BS provides unicast grants in an unsolicited manner as in UGS, but packet sizes with ertPS allocations are not fixed [2]. An MS uses its periodic allocation for both data transfer and bandwidth request adjustments (e.g., using an extended piggyback request field of the GMSH). During a silence period, the allocation is taken from the ertPS SF, and with a silenceto-talk-spurt transition, the MS sends a BR message to the BS to establish the periodic allocation during this talk spurt.
IEEE 802.16M QOS FRAMEWORK The next-generation WiMAX air interface, IEEE 802.16m advanced air interface (AAI), provides a more flexible and efficient QoS framework to support emerging and evolving mobile Internet applications. The new features
IEEE Communications Magazine • May 2010
introduced in the AAI QoS framework include a new scheduling service, adaptive granting and polling (aGP) service, quick access, delayed BR, and priority controlled access. AGP
SERVICE
In IEEE 802.16e the scheduling services UGS, ertPS, and rtPS are not efficient for applications such as online games, VoIP with adaptive multirate (AMR), and delay-sensitive TCPbased services that show ON-OFF traffic patterns with variable packet rates [4]. Furthermore, applications such as Skype show variable rate traffic patterns not only with variable packet size, but also with variable periodical intervals [5]. Therefore, it is desirable to have a more flexible QoS scheduling service to support the adaptation of both the allocation size and interarrival. A new scheduling service, aGP service [6], has been introduced in AAI to support not only granting and polling-based services, but also the adaptation of the QoS parameters to serve the dynamic traffic characteristics of applications with better efficiency. The new QoS parameters introduced in the aGP service are mandatory: primary grant polling interval (GPI) and primary grant size; and optional ones: secondary GPI, secondary grant size, and adaptation method. Advanced BS (ABS) may grant advanced MS (AMS) UL allocation GPI with grant size, or poll AMS for BR periodically every GPI. During a service, the traffic characteristics and QoS requirements may change; for example silence-suppression enabled VoIP alternates between talk-spurt and silence period, which triggers adaptation of the scheduling service state machine as described below. Adaptation of scheduling state includes switching between using primary and secondary SF QoS parameters or changing the GPI and/or grant size to values within the QoS flexibility range (i.e., without exceeding the maximal QoS requirement or violating the minimal QoS guarantee). Depending on the three adaptation methods specified during SF negotiation, the grant size and/or GPI can be changed by an ABS automatically upon detecting a certain traffic condition if the adaptation method is implicit, or triggered by explicit signaling from an AMS if the adaptation method is explicit sustained or explicit and one time only. Explicit signaling from an AMS includes a piggybacked BR, service-specific BR header, quick access message in the BR channel, or an ertPS/aGP service BR codeword in the primary fast feedback channel (P-FBCH). For explicit adaptation, if GPI_secondary and Grant_Size_secondary are defined, GPI and grant size switch between primary GPI/ Grant_Size_primary and GPI_secondary and Grant_Size_secondary as requested by the explicit signaling; otherwise, GPI and grant size changes as indicated by QoS requirements carried in the explicit signaling, as in the mechanisms mentioned above. It is important to have appropriate support when a mix of IEEE 802.16m and legacy IEEE 802.16e BSs and devices are around. An IEEE 802.16m aGP SF can be mapped to an SF of
A voice service can be served using a UGS type SF with no need for BR during a talk-spurt period; it does not need any BR during a silence period, either. However, there is a need for BR for a silence to talk-spurt. IEEE 802.16e has provisioned ertPS for this purpose.
107
ALASTI LAYOUT
4/22/10
1:57 PM
Page 108
AMS
1
DELAYED BR FOR BE
SABS
BR preamble sequence and quick access message BR ACK A-MAP IE 2 Grant for UL transmission
3
UL scheduled TX
Figure 2. Three-step random access BR procedure.
Quick access in IEEE 802.16m helps reduce the random access delay. The idea of delayed BR is to request bandwidth proactively in order to avoid random access and thus reduce the access delay. The service-specific BR header specifies a minimum grant delay to indicate the minimum delay of the requested grant for BE scheduling service. When an AMS is cleaning out its buffers, in one UL transmission it can send a delayed BR asking for future packet(s) with minimum expected grant delay if AMS can predict the future packet(s) arrival time. Hence, when the future packets do arrive, they do not need to use the lengthy random access BR procedure; instead, they can just use the dedicated UL allocation as a response to the previous delayed BR.
PRIORITY CONTROLLED ACCESS legacy IEEE 802.16e scheduling type, during AMS handover from an IEEE 802.16m network to an IEEE 802.16e network. If primary grant size value is equal to the BR header size, it means this aGP SF is primarily polling-based SF, and hence should be mapped to an rtPS SF. Otherwise, this aGP SF is primarily a grantingbased service, and thus should be mapped to an ertPS SF.
QUICK ACCESS AMS performs random-access-based BR when it has UL traffic to send but with no allocation. Random access delay is a significant part of UL access delay, which has a big impact on the end user experience. In the legacy system, the BR message is communicated from MS to BS only after random access is successful. To shorten the random access delay, the IEEE 802.16m AAI introduces a quick access message to be carried in the first step of the random access BR procedure (e.g., contention) in order to simplify the procedure from five steps in the legacy system to three steps (Fig. 2). The 12-bit station ID and 4-bit predefined BR index, which are carried in the quick access message, enable a quick exchange of BR information between ABS and AMS. The shorter random access delay can significantly improve the quality of experience (QoE) of delay-sensitive and interactive applications.
A new term, access class, is introduced to prioritize the contention-based random access. An operator can assign AMS with different access classes and block random access from certain AMSs by assigning a minimum access class of the network higher than the access class of those AMSs. The BR timer and random backoff parameters can also use different values to support differentiated random access in IEEE 802.16m.
LTE QOS FRAMEWORK The QoS level of granularity in the LTE evolved packet system (EPS) is bearer, which is a packet flow established between the packet data network gateway (PDN-GW) and the user terminal (UE or MS). The traffic running between a particular client application and a service can be differentiated into separate service data flows (SDFs). SDFs mapped to the same bearer receive a common QoS treatment (e.g., scheduling policy, queue management policy, rate shaping policy, radio link control (RLC) configuration) [3, 7]. A bearer is assigned a scalar value referred to as a QoS class identifier (QCI), which specifies the class to which the bearer belongs. QCI refers to a set of packet forwarding treatments (e.g., scheduling weights, admission thresholds, queue management thresholds, and link layer protocol configuration) preconfigured by the operator for each network element [9]. The class-based method improves
Packet filters
Packet filters Service 1
Default bearer
Service 2 Default bearer Service 3 Transport MS
LTE RAN
Gayeway
Figure 3. Default and dedicated bearers of a terminal (MS) in the LTE QoS framework.
108
IEEE Communications Magazine • May 2010
ALASTI LAYOUT
QCI
4/22/10
1:57 PM
Resource type
Page 109
Priority
Packet delay budget
Packet error loss rate
Example services
2
100 ms
10–2
Conversational voice
4
150 ms
10–3
Conversational video (live streaming)
3
3
50 ms
10–3
Real time gaming
4
5
300 ms
10–6
Non-conversational video (buffered streaming)
5
1
100 ms
10–6
IMS signaling
6
6
300 ms
10–6
Video (buffered streaming) TCP-based (e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)
7
7
100 ms
10–3
Voice Video (live streaming) Interactive gaming
8
8 300 ms
10–6
Video (buffered streaming) TCP-based (e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)
1 2 GBR
Non-GBR
9
9
Table 2. LTE standardized QCI characteristics.
the scalability of the LTE QoS framework. Figure 3 illustrates the LTE bearers. The bearer management and control in LTE follows the network-initiated QoS control paradigm, and the network initiated establishment, modification, and deletion of the bearers. LTE offers two types of bearers: • Guaranteed bit rate (GBR): Dedicated network resources related to a GBR value associated with the bearer are permanently allocated when a bearer becomes established or modified [7]. • Non-guaranteed bit rate (non-GBR): A service utilizing a non-GBR bearer may experience congestion-related packet loss. A non-GBR bearer is referred to as the default bearer, which is also used to establish IP connectivity, similar to the initial SF in WiMAX. Any additional bearer(s) is referred to as a dedicated bearer and can be GBR or non-GBR. In LTE the mapping of SDFs to a dedicated bearer is classified by IP five-tuple based packet filter either provisioned in PCRF or defined by the application layer signaling [9]. However, the default bearer typically uses a match all packet filter; any SDF that does not match any of the existing dedicated bearer packet filters is mapped onto the default bearer [9]. Therefore, if a dedicated bearer is dropped, its traffic is rerouted to the default bearer [9]. LTE specifies a number of standardized QCI values with standardized characteristics, which are preconfigured for the network elements. This ensures multivendor deployments and roaming [10]. The mapping of standardized QCI values to standardized characteristics is captured in Table 2 [10]. Besides QCI, the following are QoS attributes associated with the LTE bearer: • QCI: A scalar representing a set of packet forwarding treatments (e.g., scheduling
IEEE Communications Magazine • May 2010
weights, admission thresholds, queue management thresholds, and link layer protocol configuration). • Allocation and retention priority (ARP): A parameter used by call admission control and overload control for control plane treatment of a bearer. The call admission control uses the ARP to decide whether a bearer establishment or modification request is to be accepted or rejected. Also, the overload control uses the ARP to decide which bearer to release during overload situations [9]. • Maximum bit rate (MBR): The maximum sustained traffic rate the bearer may not exceed; only valid for GBR bearers • GBR: The minimum reserved traffic rate the network guarantees; only valid for GBR bearers • Aggregate MBR (AMBR): The total amount of bit rate of a group of non-GBR bearers In 3GPP Release 8 the MBR must be equal to the GBR, but for future 3GPP releases an MBR can be greater than a GBR. The AMBR can help an operator to differentiate between its subscribers by assigning higher values of AMBR to its higher-priority customers compared to lower-priority ones.
LTE AIR INTERFACE SCHEDULER The LTE air interface scheduler is responsible for dynamically allocating DL and UL air interface resources among the bearers appropriately while maintaining their desired QoS level in both DL and UL directions. In order to make a scheduling decision, the LTE air interface scheduler uses the following information as input: • Radio conditions at the UE identified through measurements made at the eNB and/or reported by the UE. • The state of different bearers, such as uplink
109
ALASTI LAYOUT
4/22/10
1:57 PM
WiMAX MAC was designed to support current and future QoS needs. Links can be dynamically configured based on link conditions. This dynamic configuration capability allows providing support for a range of services.
Page 110
buffer status reports (BSR) that are required to provide support for QoS-aware packet scheduling, elapsed time. • The QoS attributes of bearers and packet forwarding parameters associated with the QCIs. • The nterference situation in the neighboring cells. The LTE scheduler can try to control intercell interference on a slow basis. This improves the QoE associated with the MSs at the cell edge.
BUFFER STATUS REPORTING Similar to the bandwidth request mechanism in WiMAX, LTE also has buffer status reporting mechanism. The buffer status reporting mechanism informs the UL packet scheduler about the amount of buffered data at the UE. This mechanism consists of triggering and reporting events. The triggering event can be periodic or regular. A periodic BSR trigger does not cause a service request (SR) transmission from the UE. When a BSR event is triggered and UE has resources allocated in the physical uplink shared channel (PUSCH), then BSR is transmitted. When a regular BSR event is triggered, an SR needs to be transmitted. If SR allocation is available in the physical uplink control channel (PUCCH), the SR is transmitted at the next opportunity; otherwise, the SR is transmitted via a random access procedure. Buffer status is reported per radio bearer group. There are two BSR formats: Short and Long. Short format can be used to report on one radio bearer group whereas the long one can be used for four groups.
802.16M AND LTE COMPARISON: QOS ASPECT There are more components and functionalities in an end-to-end network providing QoS than the air interface QoS features discussed above, such as policy control and charging (PCC) functions in QoS provisioning. Here, we focus on a comparison of the QoS framework between LTE and IEEE 802.16e/IEEE 802.16m at the air interface: • QoS transport unit: The basic QoS transport unit in the IEEE 802.16e/IEEE 802.16m system is an SF, which is a unidirectional flow of packets either UL from the MS/AMS or DL packets from the BS/ABS [6]. The basic QoS transport in LTE is a bearer between UE and the PDNGW. All packets mapped to the same bearer receive the same treatment. • QoS scheduling types: There are six scheduling service types in IEEE 802.16m including UGS, ertPS, rtPS, nrtPS, and BE from IEEE 802.16e and the newly defined aGP service. LTE supports GBR and non-GBR bearers. The GBR bearer will be provided by the network with a guaranteed service rate, and its mechanism is like rtPS; the non-GBR has no such requirement and performs like BE in IEEE 802.16e/IEEE 802.16m. • QoS parameters per transport unit: Depending on the SF type, IEEE 802.16e/ IEEE 802.16m can control maximum packet delay and jitter, maximum sustained traf-
110
fic rate (MSTR), minimum reserved traffic rate (MRTR), and traffic priority. LTE MBR and GBR are similar to IEEE 802.16e/IEEE 802.16m MSTR and MRTR, respectively. However, MBR and GBR are only attributes of GBR bearers, while in IEEE 802.16e/IEEE 802.16m even a BE SF can be rate limited using its MSTR. Also, with 3GPP Release 8, GBR and MBR are set equal, while IEEE 802.16e/IEEE 802.16m allows the operator to select independent values for MSTR and MRTR. On the other hand, LTE AMBR allows the operator to rate cap the total non-GBR bearers of a subscriber. • QoS handling in the control plane: The SF QoS parameters are signaled in IEEE 802.16e/IEEE 802.16m via DSx/AAI-DSx messages. In LTE the QCI and associated nine standardized characteristics are not signaled on any interface. Network initiated or client initiated QoS are both supported in IEEE 802.16e/IEEE 802.16m systems. Therefore, both operator managed service and unmanaged service can be supported. The flexible architecture gives the mobile client opportunities for differentiation. LTE only supports network initiated QoS control. • QoS user plane treatment: The ARP parameter in LTE provides the following flexibilities to the operator: –Accept or reject establishment or modification of bearers during the call admission control decision based on not only the requested bandwidth, available bandwidth, or number of established bearers, but also the priority of the bearer –Selectively tear down bearers based on their priorities during an overload situation
CONCLUSIONS Providing the required QoS is vital to deliver a good user experience over mobile Internet. The notion of QoS is becoming even more important as device capabilities have revealed the desire for consumers to use more rich media content such as video. Fourth-generation wireless technologies such as IEEE 802.16e, IEEE 802.16m, and LTE are designed to support current and future QoS needs. Connection-oriented per-flow-based unidirectional QoS support in IEEE 802.16e allows several service flow types such as UGS, rtPS, nrtPS, ertPS, and BE to deliver real-time and non-real-time traffic. The UL bandwidth request and granting mechanism allows MSs to request and receive the required resources to transmit data in the UL direction. Advanced features such as a new scheduling service (i.e., aGP), quick access, and delayed bandwidth request in IEEE 802.16m further enhances the capabilities in providing the required QoS for next-generation mobile Internet applications. The LTE QoS mechanisms follow a network initiated QoS control based on GBR and non-GBR bearers, which is a class-based packet forwarding treatment for delivering real-time and non-real-time traffic. This article explains the QoS framework of IEEE 802.16e, IEEE 802.16m, and LTE, and compares their QoS features against each other.
IEEE Communications Magazine • May 2010
ALASTI LAYOUT
4/22/10
1:57 PM
Page 111
REFERENCES
BIOGRAPHIES
[1] Keynote, “Rich Internet Applications: Design, Measurement, and Management Challenges,” White Paper, 2009. [2] IEEE Std 802.16-2009, “Part 16: Air Interface for Broadband Wireless Access Systems.” [3] 3GPP TS 23.401 v. 8.8.0, “General Packet Radio Service (GPRS) Enhancements for Evolved Universal Terrestrial Radio Access Network (E-UTRAN) Access,” Dec. 2009. [4] C. Nie, M. Venkatachalam, and X. Yang, “Adaptive Polling Service for Next-Generation IEEE 802.16 WiMAX Networks,” IEEE GLOBECOM, 2007. [5] J. Zhu, “On Skype Voice Traffic Characteristics, Rate Adaptation, and User Experience,” to appear, IEEE WCNC 2010, Apr. 2010. [6] IEEE 802.16m/D4, “Draft Amendment to IEEE Standard for Local and Metropolitan Area Networks — Part 16: Air Interface for Fixed and Mobile Broadband Wireless Access Systems, Advanced Air Interface,” Feb. 2010. [7] 3GPP TS 36.300 v. 8.11.0, “Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access (E-UTRAN); Overall Description; Stage 2,” Jan. 2010. [8] M. Alasti and B. Neekzad,”WiMAX QoS,” Ch. 10, WiMAX Technology and Network Evolution, K. Etemad and M. Lai, Eds., Wiley-IEEE Press, 2010. [9] H. Ekstrom et al., “QoS Control in the 3GPP Evolved Packet System,” IEEE Commun. Mag., Feb. 2009, pp. 76–83. [10] 3GPP TS 23.203 v. 8.8.0, “Policy and Charging Control Architecture,” Dec. 2009.
M EHDI A LASTI received a Ph.D. in electrical and computer engineering from the University of Maryland at College Park in 2001. He has been with Clearwire as a wireless technical consultant working in different 4G wireless technologies such as WiMAX and LTE. Prior to Clearwire he was with Nextwave, Sprint Nextel, Airvana, and Zagros Networks.
ADDITIONAL READING [1] IEEE Std. 802.16-2009, “Standard for Local and Metropolitan Area Networks — Part 16: Air Interface for Broadband Wireless Access Systems,” May 2009.
IEEE Communications Magazine • May 2010
B EHNAM N EEKZAD received his Ph.D. in electrical and computer engineering from the University of Maryland at College Park. He has been involved in different aspects of 4G wireless technologies development and network architecture engineering, including end-to-end QoS architecture design. He is currently with Clearwire and prior to that was with Sprint Nextel. JIE HUI received her B.S. in electrical engineering from Xi’an Jiaotong University in 1996, and her M.S. in electrical engineering from the Institute of Space Electronic Equipment in 1999 in China. She was with ZTE Corp. from 2000 to 2001. After graduation from North Carolina State University with a Ph.D. in CPE, she joined Intel Labs in 2006. Her research interests include QoS provisioning and QoE optimization for multimedia over wireless networks, medium access control, modeling, and performance evaluations of wireless networks. RATH VANNITHAMBY [SM] received his B.S., M.S., and Ph.D. in electrical engineering from the University of Toronto, Canada. He is currently a senior research scientist at Intel Labs. He leads and manages a team responsible for 4G/WiMAX research and standardization. Previously, he was a researcher at Ericsson. He has published 25+ papers and has 75+ patents granted/pending. He has authored chapters of two books. He was a TPC member for ICC, GC, VTC, and WCNC, and a Guest Editor for EURASIP Journal on Wireless Communications and Networking. His research interests are in the area of MAC, RRM, and cross-layer optimization for 4G.
111