1
`
`Congestion Control in WCDMA with Respect to Different Service Classes
`
`Joachim Sachs, Thomas Balon, Michael Meyer
`
`Ericsson Eurolab Deutschland GmbH
`Ericsson Allee 1, 52134 Herzogenrath, Germany
`e-mail: {Joachim.Sachs | Michael.Meyer}@eed.ericsson.se
`
`Universal Mobile
`the
`In
`–
`Abstract
`(UMTS)
`the radio
`Telecommunications System
`resource management functions are jointly handled in
`two different protocol layers, the Radio Resource
`Control (RRC) and the Medium Access Control
`(MAC). Congestion control functions are performed
`in the RRC layer. Therefore congestion control
`requires strong
`interactions between
`the radio
`interface protocol
`layers,
`including measurement
`reports which are
`transmitted, as well as
`reconfiguration procedures.
`In a simulation environment different radio resource
`reconfiguration procedures are
`evaluated
`for
`congestion control. One aspect is the method which is
`used to detect congestion. The system performance is
`presented for a mix of traffic with different service
`classes.
`It is demonstrated how during a congestion, different
`service classes can be differentiated. This provides
`improved quality of service provisioning for the users
`according to their service requirements. By choosing
`the best congestion indicator, a good trade-off can be
`found between quick congestion resolution and
`quality of service control on one hand, and efficient
`utilization of the radio resources on the other hand.
`
`I. INTRODUCTION
`
`The Universal Mobile Telecommunications System
`the European 3rd generation mobile
`(UMTS)
`is
`telecommunication system. In a global co-operation of
`standards organisations it is being standardised at the
`Third Generation Partnership Project (3GPP), where a
`technical specification is being developed on the basis of
`an evolved GSM core network and the UMTS Terrestrial
`Radio Access (UTRA). The resulting specification is
`contributed
`to
`ITU
`for
`the
`International Mobile
`Telecommunications 2000 (IMT-2000) standards family.
`Requirements for UMTS are to support multimedia
`services with data rates of up to 384 kbit/s for wide-area
`coverage and up to 2 Mbit/s for indoor and low-range
`outdoor coverage. Furthermore, UMTS has to provide a
`high service flexibility to support both circuit- and
`packet-switched services with a wide range of applicable
`data rates. Additionally, it must be possible to use
`multiple services simultaneously. Hence, UMTS will
`make many data services available to mobile users,
`especially those currently known from the Internet.
`The paired frequency bands of UMTS – 1920-1980
`MHz in uplink and 2110-2170 MHz in downlink – use
`the Frequency Division Duplex (FDD) mode which is
`based on Wideband Code Division Multiple Access
`
`(WCDMA) technology. In contrast to Time Division
`Multiple Access technology, the radio resources in
`WCDMA are not as easily “countable”. This property is
`also referred to as soft-capacity, meaning that the balance
`of capacity, quality and coverage can be shifted towards
`improving any of these characteristics for the price of
`reducing the other ones. For this reason, congestion
`control mechanisms in WCDMA are challenging to
`define, since there are no hard limits imposed by the
`system owing to the soft-capacity property.
`In this paper different congestion control mechanisms
`will be discussed and analysed. This comprises the
`methods of how congestion is detected, how radio
`resources are reassigned and which protocol sequences
`are involved.
`In section II the UMTS Terrestrial Radio Access
`Network (UTRAN) is described, including the properties
`of the WCDMA channel and the protocols which are
`involved in the congestion control process. In section III
`a simulation scenario
`is presented and results on
`congestion control mechanisms are provided in section
`IV. Eventually, a conclusion is drawn in section V.
`
`II. UTRAN
`
`A. WCDMA Channel
`
`The physical layer that is used in the paired frequency
`bands of UMTS is based on WCDMA [1]. Variable data
`rates can be provided on the transport channels by using
`codes with different spreading factors. This implies that
`transport channels with higher data rates have a reduced
`spreading gain. In order to balance the reduced spreading
`gain and still achieve the required ratio of received
`energy per bit to the effective noise power, Eb/N0, the
`output power level has to be increased. On dedicated
`channels, this is achieved by fast power control. 1600
`power control commands per second are transmitted from
`the receiver to the transmitter, so that the transmitter can
`adapt its transmit power accordingly. When multiple
`users are transmitting at the same time, each user
`increases the interference level for all other users.
`Therefore the radio link quality depends not only on the
`radio channel attenuation, but also on the traffic load
`caused by all mobile
`terminals
`in
`the same and
`neighbouring cells. In this sense, the common shared
`resource is power. An increase in interference forces each
`mobile terminal to also increase the transmit power. This
`in turn increases the interference for other users and they
`will react in the same way. In general, this converges to a
`certain transmit power for all users. However, when the
`traffic load becomes too high this can lead to a
`
`Wireless’99 | 4. ITG-Fachtagung "Mobile Kommunikation"
`October 6 - 8, 1999, Munich, Germany
`
`

`

`Measurements
`
`5.
`
`Control
`
`Control
`
`Measurements
`
`Control
`
`R R C
`
`R L C
`
`M A C
`
`L 1
`
`3.
`
`R R C
`
`Measurement Report
`
`4.
`
`Radio Resource
`Assignment
`
`6.
`
`RLC retransmission
`control
`
`R L C
`
`M A C
`
`1.
`
`L 1
`
`Control
`
`5.
`
`Control
`
`Measurements
`
`Control
`
`Measurements (Interference Level)
`
`2.
`
`U T R A N
`U E
` Figure 1 Radio interface protocol reconfiguration
`
`1) L1 – Physical Layer (PHY)
`
`The tasks of the physical layer comprise to perform
`forward error correction (FEC), error detection and
`interleaving. Further
`tasks are
`the multiplexing of
`transport channels, spreading, modulation and radio
`frequency processing. The physical
`layer
`is also
`responsible for frequency and time synchronisation and
`closed-loop power
`control. The power
`control
`dynamically adapts the transmitter power to the channel
`conditions, so that the receiver always receives the signal
`at the correct power level. For that the receiver informs
`the transmitter with power control commands at a rate of
`1600 signals per second. Another function of the physical
`layer is to measure transmission values, e.g. transmit
`power, interference power, signal-to-interference ratio
`(SIR), block error rate. These values are indicated to the
`Radio Resource Control layer, which uses them for radio
`resource management.
`For the physical layer, a set of Transport Formats (TF)
`– the so-called Transport Format Set (TFS) – is assigned
`to each transport channel. A Transport Format can be
`described as a combination of channel encoding,
`interleaving, bit rate and mapping onto a physical
`channel. The Transport Format Set specifies those
`Transport Formats which a radio access bearer can use on
`a transport channel. The physical layer multiplexes one
`or more transport channels onto physical channels. When
`transport channels are multiplexed not all possible
`combinations of Transport Formats of the different
`transport channels may be applied by the physical layer
`at a certain point in time. An authorised combination of
`Transport Formats that may be applied simultaneously is
`referred to as a Transport Format Combination (TFC).
`The set of Transport Format Combinations that can be
`applied by the physical layer is called Transport Format
`Combination Set (TFCS). It is a subset of all possible
`Transport Format Combinations.
`For example, a transport channel 1 (TC1) and transport
`channel 2 (TC2) shall have the following Transport
`Format Sets:
`‰ TC1: TFS1 = {0, 8, 16, 32} kbit/s,
`‰ TC2: TFS2 = {16, 32} kbit/s.
`Then a resulting TFCS could be:
`‰ TFCS = {16, 24, 32, 40} kbit/s,
`
`2
`
`congestion situation. If some users reach the limit of their
`transmit power they cannot follow the power control
`process any longer. Therefore they perceive a lower
`Eb/N0 and an increased error rate on the channel.
`
`B. UMTS Radio Access Bearers
`
`The UTRAN provides radio access bearers for data
`services. They offer a bearer transport between the
`mobile user equipment and the core network. These radio
`access bearer services can be configured very flexibly in
`order to support the quality of service requirements of a
`data service. A given radio access bearer is defined by a
`set of parameters, like maximum bit rate, bit error rate
`and transmission delay.
`Four quality of service classes are defined for UMTS:
`‰ Conversational Class,
`‰ Streaming Class,
`‰
`Interactive Class,
`‰ Background Class.
`
`The Conversational Class is used for conversational
`real time applications, like e.g. voice over IP or video
`conferencing. Requirements are a low transmission
`delay, a low jitter and a low round trip delay.
`The Streaming Class is used for real time streaming
`applications, e.g. audio or video streaming. The
`requirements on the transmission delay are less stringent
`then for the Conversational Class. However, the timing
`relations between data packets have to be preserved and
`therefore the jitter must be low.
`The Interactive Class is used for best effort data
`services which are following a request-response pattern.
`Examples for such services are WWW traffic or telnet.
`Requirements are a low residual error rate and a low
`round trip delay.
`The Background Class is used for best effort data
`services with no requirements on delay. Examples are file
`transfer and e-mail.
` For each radio access bearer a link is established and
`all
`radio
`interface protocols are configured with
`appropriate parameters.
`
`C. UTRAN Radio Interface Protocols
`
`The UTRAN radio interface is layered in three
`protocol layers:
`‰
`the physical layer
`‰
`the data link layer
`‰
`the network layer
`
`(L1),
`(L2),
`(L3).
`
`interface protocol
`
`the radio
`The relationship of
`architecture is depicted in Fig. 1.
`The radio resource management functions are split
`into two parts which are handled in two protocol layers.
`The Radio Resource Control (RRC) layer is responsible
`for assigning a pool of radio resources to each radio
`access bearer and reconfiguring the radio resources. The
`Medium Access Control (MAC) manages this pool of
`radio resources instantaneously. So to speak, MAC has
`short-term control over the radio resources and RRC
`defines the long-term guidelines for MAC.
`
`Wireless’99 | 4. ITG-Fachtagung "Mobile Kommunikation"
`October 6 - 8, 1999, Munich, Germany
`
`

`

`thus excluding the combinations of using 32|16 kbit/s and
`32|32 kbit/s.
`The physical layer is configured by RRC when a radio
`access bearer is setup or reconfigured. The TFSs and
`TFCS are then assigned by RRC. A detailed description
`of the physical layer can be found in [1] and [6].
`
`2) L2 – Medium Access Control Protocol (MAC)
`
`The MAC layer is a sublayer of the data link layer. It
`provides logical channels to the RLC layer. Logical
`channels describe “what
`type” of
`information
`is
`transported. The MAC layer maps logical channels onto
`transport channels. Transport channels are provided by
`the physical layer to the MAC layer, and they describe
`“how” information is transported.
`the
`for
`Furthermore, MAC
`is
`responsible
`is
`instantaneous radio resource management. This
`achieved by selecting the Transport Format for each
`transport channel, taking into account the instantaneous
`source rate. In this process MAC also has to prioritise
`between data flows of the same user equipment. MAC is
`limited to the TFSs and TFCS which have been assigned
`to it by RRC. MAC multiplexes packet data units from
`RLC onto the transport channels. The instantaneous radio
`resource management is handled every transmission time
`interval, e.g. 10 ms.
`is responsible for managing
`Furthermore, MAC
`common transport channels, which are not considered in
`the analysis of this paper. For further information refer to
`[5].
`
`3) L2 – Radio Link Control Protocol (RLC)
`
`The RLC layer is a sublayer of the data link layer.
`There is one RLC connection per radio access bearer.
`Higher layer data is segmented into packet data units of
`sizes which are suitable for the radio transmission. On
`the receiving side the original data format is reassembled.
`The RLC layer provides three different data transfer
`modes to the higher layer.
`The acknowledged data transfer mode is a guaranteed
`service. All higher layer data is transferred error-free
`over
`the
`radio
`interface. This
`is achieved by
`retransmissions of erroneous RLC packet data units. This
`mechanism is also referred to as automatic repeat request
`(ARQ). An efficient radio access bearer configuration
`can be achieved by finding a good balance between
`forward error correction in the physical layer and ARQ in
`the link layer.
`no
`transfer mode,
`data
`In unacknowledged
`retransmissions are performed. The transmission delay is
`reduced, but the delivery of higher layer data cannot be
`guaranteed.
`In transparent data transfer mode, no specific protocol
`information is added to the packet data units and
`therefore
`the protocol overhead
`is minimized. No
`retransmissions are performed.
`The RLC connection is configured by RRC when a
`radio access bearer is setup or reconfigured [4].
`
`3
`
`4) L3 – Radio Resource Control Protocol (RRC)
`
`The RRC protocol is located in the control plane of the
`network layer. It handles the control plane signalling
`between the user equipment (UE) and the core network.
`It is also responsible for the management of radio
`resources, comprising establishment, reconfiguration and
`release of radio access bearers. It is the task of RRC to
`control the requested quality of service and perform
`congestion control.
`Several radio access bearers can be assigned to the
`same user equipment. The task of RRC is to manage the
`available radio resources among all radio access bearers
`of all mobile terminals within the cell. The radio resource
`management procedures are performed in the RRC entity
`on the UTRAN side of the radio link. The appropriate
`information is then transmitted to the RRC entities in the
`mobile terminals. The other radio link protocols are then
`configured accordingly by the RRC entities.
`RRC controls the parameters and settings of the
`underlying protocol layers RLC, MAC and PHY, thus
`selecting the radio access bearer processing [3]. It signals
`to MAC and PHY the Transport Format Set and
`Transport Format Combination Set for the transport
`channels. Additionally, the RLC data transfer mode is
`assigned.
`
`D. Congestion Control
`
`In contrast to other mobile communication systems,
`the radio network controller
`in UMTS does not
`permanently control the actual uplink data rates of the
`mobile terminals. As described in the previous section,
`the RRC in the network assigns a pool of radio resources
`to the radio access bearers and MAC manages these
`resources instantaneously. Thus, the mobile stations
`choose their data rates – within the limitations given by
`RRC – independent from the radio network controller.
`Due to the movement of the mobile stations, the
`characteristics of the radio channel and the traffic
`transmitted by all applications, overload situations may
`occur in a system. In such a situation the interference in
`the system increases rapidly, leading to a reduced quality
`of service or call dropping.
`A fast detection of an overload situation can be based
`on several measurements, e.g.:
`‰
`in the sender by monitoring the transmit power and
`the power control commands,
`in the receiver by monitoring the interference power,
`the signal-to-interference ratio or the block error
`rate.
`
`‰
`
`The congestion can be resolved by reducing the
`interference level. This can be achieved by reducing the
`allowed data rates of one or several radio access bearers.
`As stated before, the radio network controller in UMTS
`does not permanently control the data rates of the radio
`access bearers. Therefore several steps are necessary to
`react to congestion. First the congestion needs to be
`detected by measurements, which have to be signalled to
`the RRC layer. If the RRC entity in the UTRAN
`recognises congestion it reconfigures the radio resources
`
`Wireless’99 | 4. ITG-Fachtagung "Mobile Kommunikation"
`October 6 - 8, 1999, Munich, Germany
`
`

`

`in order to resolve the congestion. There are several
`possibilities to reconfigure the radio resources for one or
`more mobile terminals [3]:
`‰ Limiting the TFCS of the transport channels. This is
`achieved by sending the RRC message Transport
`Format Combination Control from the UTRAN to
`the user equipment. The TFCs in the TFCS can thus
`be temporarily restricted, excluding too high data
`rates. The advantage of that method is that the
`message does not need to be acknowledged from the
`user equipment, since all TFCs are still valid and the
`message size is comparatively small. Even if the
`message is lost, the unlimited TFCS can still be
`used. However, this method is not as flexible as
`when a completely new TFCS is used.
`‰ Assigning a completely new TFCS to a user
`equipment. This is achieved by sending the RRC
`message Transport Channel Reconfiguration from
`the UTRAN to the user equipment, containing the
`new TFCS. The message needs to be confirmed by a
`Transport Channel Reconfiguration Complete
`message, before the new TFCS becomes valid.
`‰ Reconfiguring the whole radio access bearer. In
`difference to the previously described transport
`channel reconfiguration, this method includes a
`complete reconfiguration of all the protocol entities
`of the radio access bearer. The RRC message Radio
`Access Bearer Reconfiguration is send from the
`UTRAN to user equipment. The message needs to be
`confirmed
`by
`a
`Radio
`Access
`Bearer
`Reconfiguration Complete message, before the new
`radio access bearer becomes valid.
`If the congestion detection is based on a measurement
`from the user equipment, a measurement report has to be
`transmitted from the RRC entity in the user equipment to
`the RRC entity in the UTRAN prior to any of the above
`mentioned congestion control schemes.
`In the presence of radio access bearers with different
`quality of service configurations, the congestion control
`is performed according to the service class. For example,
`only radio access bearers for Background Class traffic
`and Interactive Class traffic are reconfigured, while real
`time services keep their configuration.
`
`III. SIMULATION SCENARIO
`
`the
`to analyse
`in order
`is used
`A simulator
`performance of congestion control algorithms in UMTS.
`The simulator comprises all protocols of the radio
`interface, RRC, RLC, MAC, a WCDMA channel model
`and traffic sources generating e.g. IP traffic. The UMTS
`core network and external packet data networks are not
`considered. Objective of the simulations is to analyse
`congestion control methods, including the influence of
`the protocols which are involved. Hereby users with
`different types of applications and thus different service
`requirements are assumed.
`for connection
`responsible
`The RRC
`layer
`is
`establishment, connection release and the radio access
`bearer reconfiguration in case of congestion. It handles
`measurement reports from
`the physical
`layer and
`transmits them, if necessary, from the user equipment to
`
`4
`
`the UTRAN. During connection establishment RRC
`assigns a Transport Format Combination Set to the user
`equipment according
`to
`the
`requirements of
`the
`application. The congestion control mechanism triggers
`the reconfiguration of the transport channel when a
`congestion situation occurs. This reconfiguration is done
`by assigning new TFCSs to the mobile terminals as
`described
`in
`the
`previous
`section. After
`the
`reconfiguration, the reduced data rate leads to a lower
`interference level.
`RLC is responsible for the segmentation of higher
`layer datagrams into RLC packet data units of size 10
`bytes on the transmitting side and the corresponding
`reassembly at the receiving side. A selective repeat ARQ
`mechanism is used.
`The MAC layer selects the Transport Format Set out
`of the Transport Format Combination Set according to
`the RLC queue length.
`The physical layer includes a WCDMA channel model
`and the signalling of measurement reports to the RRC
`layer. The WCDMA channel model calculates a Eb/N0
`value at the receiver for each transport channel. The
`Eb/N0 is then mapped to a block error rate, which has
`been derived from dedicated physical layer simulations.
`These simulations include the effects of multipath fading
`and closed loop power control, which are therefore
`included in the block error rate. The calculation of the
`Eb/N0 is based on the received signal power, the
`spreading factor and the interference caused by all other
`active mobile terminals. In an iterative process, the
`transmit power of all active mobile terminals is adapted
`until each reaches its target Eb/N0 unless this is prohibited
`by the transmit power range. This process is performed
`once at the beginning of each transmission time interval
`of 10 ms and the Eb/N0 is then assumed to be constant
`over that transmission time interval. The target Eb/N0 is
`chosen as 3 dB corresponding to a block error probability
`of roughly 13% and the transmit power is upper bound
`by 27 dBm. The delay for physical layer processing and
`transmission has been set to 15 ms.
`For the simulations a single cell system is assumed
`with a certain number of active mobile terminals which
`are initially randomly distributed within the cell. Two
`classes of mobile terminals are distinguished according to
`their type of application:
`
`Class A: This type of UE has a single delay sensitive
`real-time connection. The Transport Format is
`set to a data rate of 16 kbit/s. The generated
`traffic consists of constant rate IP packets with
`a packet size of 100 bytes. This type of traffic
`is generated e.g. by a voice over IP traffic
`source.
`Class B: This type of UE has a single best effort
`connection that is not delay sensitive, e.g. a
`WWW session. The Transport Format Set is
`chosen to allow variable data rates between 8
`and 32 kbit/s. The IP packet size is 1000 bytes.
`In the simulation scenario a mixture of mobile
`terminals of both classes are assumed, where 25% of the
`terminals are of class A and the remaining 75% of class
`
`Wireless’99 | 4. ITG-Fachtagung "Mobile Kommunikation"
`October 6 - 8, 1999, Munich, Germany
`
`

`

`5
`
`cell a degradation of the service is perceived. The reason
`for the degradation is, that in a growing number of cases
`the mobile terminals are not able to transmit at the power
`which is necessary to achieve the required Eb/N0 against
`the
`larger
`interference
`introduced by other users.
`Consequently, the block error rate increases which leads
`to a higher number of retransmissions in the RLC layer
`and the IP throughput decreases.
`In contrast to the mobile terminals of class A, those of
`class B see a degradation of their mean IP throughput at a
`much lower load. This is depicted in Fig. 3 for the same
`interference thresholds as in Fig. 2. With increasing load
`an increasing number of mobile terminals with best effort
`service are reconfigured to lower data rates. The
`paradigm of the applied method is to provide to class A
`users the required service and fill the remaining capacity
`with traffic of class B. However the service quality of
`users of class A shall not be degraded owing to the traffic
`of class B. From Fig. 2 and 3 it becomes obvious that this
`paradigm works better, if a reconfiguration of the data
`rate of the mobile terminals of class B is triggered at a
`lower interference threshold. However, the total cell
`capacity is not utilised as efficiently as with higher
`interference thresholds. This can be seen in Fig. 4, where
`the accumulated IP traffic of the users of class A and B is
`depicted. With an interference threshold of –93 dBm the
`total throughput in the cell is up to 15% increased
`compared to an interference threshold of –95 dBm. This
`can be explained by the effect that reconfiguration is
`performed more conservatively. On the other hand, at a
`very high traffic load a high interference threshold can be
`less stable as can be seen in Fig. 4 for 20 mobile
`terminals on average in the cell. At this point the
`interference is already very high, and the resulting high
`block error rate delays the reconfiguration messages
`which are transmitted.
`In Fig. 5 a comparison is drawn between the different
`options of reconfiguring the radio resources, which were
`presented in section II D. The reconfiguration delay
`specifies the time interval between the moment that
`congestion has been detected until the radio resources
`have been reconfigured. If a radio access bearer or
`transport channel is reconfigured, a RRC message is
`transmitted from the UTRAN to the UE and
`
`Interference limit = -95 dBm
`
`Interference limit = -94 dBm
`
`Interference limit = -93 dBm
`
`35
`
`30
`
`25
`
`20
`
`15
`
`10
`
`5
`
`0
`
`Mean uplink IP throughput per mobile station [kbit/s]
`
`B. The session arrival and session duration are
`exponentially distributed and the call duration has a mean
`value of 120 s.
`In the scenario as it has been chosen, the two service
`classes A and B are differentiated in the congestion
`control mechanism. The users of class A have only one
`TF in the TFS that provides a single data rate of 16 kbit/s,
`so the data rate for those users cannot be reduced. For the
`users of class B the TFS contains TFs for several data
`rates. In this way, the users in class A – having a harder
`service requirement – can maintain their service level for
`a higher traffic load by reducing the quality of service for
`the users in class B. Nevertheless, during the time it takes
`to resolve the congestion the quality of service of the
`mobile terminals of class A is also reduced. Therefore a
`fast and efficient congestion control mechanism is
`required to achieve a high cell capacity, while still
`maintaining a good quality of service according to the
`service class.
`Note, that in this example, no priority scheduling in
`the MAC layer can be applied for service differentiation
`of uplink traffic, since each mobile terminal has only one
`single data connection of either class A or B. In
`downlink, the same network node is responsible for the
`transmission to all mobile terminals within a certain area
`and thus the knowledge about the traffic volume and
`service classes can be exploited for scheduling.
`
`IV. SIMULATION RESULTS
`
`For the simulation results presented in Fig. 2-4 a
`congestion control mechanism is applied comprising the
`following steps, as indicated in Fig. 1: In the UTRAN the
`interference level at the receiver is measured (1). When a
`certain threshold is surpassed a measurement report is
`signalled to the RRC entity (2). RRC makes a decision on
`reassigning new TFCSs for one or more mobile terminals
`(3) which are then sent to the user equipment (4). On
`UTRAN and UE side, the radio bearers are reconfigured
`accordingly (5) and a confirmation is sent (6).
`The mean IP throughput for mobile terminals of class A
`versus the number of active mobile terminals is depicted
`in Fig. 2 for different interference thresholds. It can be
`seen that up to a mean load of approximately 12 mobile
`terminals all users of class A can keep their requested
`data rate and only with still further increasing load in the
`18
`
`Interference limit = -95 dBm
`
`Interference limit = -94 dBm
`
`Interference limit = -93 dBm
`
`16
`
`14
`
`12
`
`10
`
`02468
`
`Mean uplink IP throughput per mobile station [kbit/s]
`
`2
`
`4
`
`6
`
`12
`10
`8
`Average number of mobile terminals
` in cell
` Figure 2 Mean application data rate per UE for class A
`
`14
`
`16
`
`18
`
`20
`
`2
`
`4
`
`6
`
`14
`12
`10
`8
`Average number of mobile terminals in cell
`
`16
`
`18
`
`20
`
` Figure 3 Mean application data rate per UE for class B
`
`Wireless’99 | 4. ITG-Fachtagung "Mobile Kommunikation"
`October 6 - 8, 1999, Munich, Germany
`
`

`

`UE triggered bearer reconfiguration
`
`UTRAN triggered bearer reconfiguration
`
`UTRAN triggered TFCS limitation
`
`250
`
`200
`
`150
`
`100
`
`50
`
`0
`
`Uplink reconfiguration delay [ms]
`
`6
`
`2
`
`4
`
`6
`
`14
`12
`10
`8
`Average number of mobile terminals in cell
` Figure 5 Mean application data rate per UE for class B
`that an extended TFCI field can be used, increasing the
`size of the TFCS to 1024 [6]. This reduces the probability
`of having too few TFCI values to apply TFCS limitation.
`
`16
`
`18
`
`20
`
`V. CONCLUSION
`
`The impact of the WCDMA radio channel of the
`UMTS FDD mode on radio resource usage has been
`described. Congestion control methods have been
`discussed with different radio resource reconfiguration
`mechanisms and congestion detection indicators.
`The choice of an optimal threshold for the congestion
`indication measure allows to find the appropriate balance
`between efficient utilisation of the cell capacity on one
`hand and quickly resolving congestion and maintaining
`the quality of service for the data services on the other
`hand.
`Simulations have shown how users of different service
`classes can share the radio resources according to their
`service requirements. Results indicated that it is possible
`to shield traffic flows with strict service requirements
`from best effort flows. Radio resources provided to best
`effort service traffic can dynamically be adapted to use
`only resources not required by traffic with strict service
`requirements.
`
`VI. REFERENCES
`[1] E. Dahlman, P. Beming, J. Knutsson, F. Ovesjö, M.
`Persson, and C. Roobol, “WCDMA – The Radio
`Interface
`for
`Future Mobile Multimedia
`Communications,” IEEE Transactions on Vehicular
`Technology, Nov. 1998.
`[2] 3GPP, TSG RAN, WG 2, Services provided by the
`Physical Layer, TS 25.302 V2.3.0, June 1999.
`[3] 3GPP, TSG RAN, WG 2, Radio Resource Control
`(RRC) Protocol Specification, TS 25.331 V1.1.0,
`June 1999.
`[4] 3GPP, TSG RAN, WG 2, Radio Link Control
`(RLC) Protocol Specification, TS 25.322 V1.1.0,
`June 1999.
`[5] 3GPP, TSG RAN, WG 2, Medium Access Control
`(MAC) Protocol Specification, TS 25.321 V3.0.0,
`June 1999.
`[6] 3GPP, TSG RAN, WG 1, Multiplexing and channel
`coding (FDD), TS 25.212 V2.0.0, June 1999.
`
`Interfere nce Lim it = -9 5 dBm
`
`Interfere nce Lim it = -9 4 dBm
`
`Interfere nce Lim it = -9 3 dBm
`
`200
`
`180
`
`160
`
`140
`
`120
`
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`Total uplink IP throughput of the cell [kbit/s]
`
`18
`
`20
`
`2
`
`4
`
`16
`14
`12
`10
`8
`6
`Average num ber of m obile term inals in cell
` Figure 4 Mean application data rate per UE for class B
`is acknowledged with a complete message, thus resulting
`in two messages which are transmitted over the radio
`link. In the simulations the size of the reconfiguration
`messages has been set
`to 30 bytes
`leading
`to
`segmentation into several RLC packet data units. If the
`congestion detection is based on measurements in the
`mobile terminal, an additional measurement report has to
`be sent to the UTRAN prior to the reconfiguration. An
`example for that is when the detection for uplink
`congestion is based on transmit power information of the
`physical layer in the mobile terminal. Therefore a third
`message over the radio link is introduced, thus adding to
`the reconfiguration delay. The reconfiguration delay is a
`significant measure for the congestion control, since
`during that time interval a congestion is leading to an
`increased block error probability on the radio link and
`thus an increased number of RLC retransmissions. This
`effect is depicted in Fig. 5. For a higher traffic load in the
`cell, the reconfiguration delay increases due to the
`increased number of retransmissions.
`Another option to handle congestion, is the limitation
`of the valid TFCS by means of an RRC message
`Transport Format Combination Control. This is a short
`message, which needs to be transmitted only from the
`UTRAN to the mobile terminal. In Fig. 5 it is shown that
`the reconfiguration delay for TFCS limitation is much
`shorter compared to a transport channel reconfiguration.
`It is also less depending on the traffic load owing to its
`small message size. A large reconfiguration message is
`segmented into several RLC datagrams. Therefore it is
`immensely stronger influenced by increased block error
`rates. However, the TFCS limitation has the drawback of
`low
`flexibility. The momentary Transport Format
`Combination used in a frame on the physical channel is
`indicated in an identity field (TFCI) of 6 bits, providing a
`maximum
`of
`64
`different Transport
`Format
`Combinations. If several parallel services are active and
`service k has lk different Transport Formats, then the
`number of possible TFCs can be rather large: (cid:213)
`kl
`.
`
`k
`Format
`of Transport
`number
`the
`Therefore,
`Combinations is limited and must be used efficiently. A
`TFCS limitation reduces the number of TFCs that can be
`used whereas a traffic channel reconfiguration reassigns a
`new TFCS and is therefore more flexible. In the
`meantime it has been added to the UMTS specification
`
`Wireless’99 | 4. ITG-Fachtagung "Mobile Kommunikation"
`October 6 - 8, 1999, Munic