The radio link control (RLC) protocol is implemented between RNC and UE entities in the UTRAN. The Acknowledge Mode (AM) RLC protocol provides a reliable data transmission in the HSPA network for the best effort traffic. The packet losses over the Uu and the Iub interface need to be recovered using the RLC protocol and that avoids long retransmission delays by TCP which is located at the end entities. In this way, a minimum loss rate for the best effort traffic in the UTRAN network can be guaranteed and the required number of TCP retransmissions by the end entities is also minimised. Any retransmissions at TCP level do not only cause additional delay but also consume additional network resources. Further, by hiding packet losses at the UTRAN network, TCP can provide effective flow control between end nodes to maximise the throughput and optimise the network utilisation. In short, having RLC AM mode for best effort traffic within the UTRAN enhances the overall end-to-end performance. The following section briefly describes the RLC protocol implementation within the HSPA network simulator.
3.6.1
Overview of RLC protocol
The RLC protocol does not only provide the AM mode operation but also two other modes: Transparent Mode (TM) and Unacknowledged Mode (UM). In TM mode, data will be sent to the lower layer without adding any header information. In UM mode, data will be sent to the lower layer with the header information, however similar to UDP, there is no recovery scheme. Hence it is recommended to use this mode of operation for most of the real time services which do not require strict constraints for the loss but which have critical limits for the delay. In AM mode, as described above, the RLC protocol works like TCP. In this mode of operation, a reliable data
transmission is guaranteed between peer RLC entities by using an ARQ based recovery scheme. Further details about the RLC protocol are given in the 3GPP specification [22].
The RLC protocol is implemented in the HSPA network simulator. The main focus of this implementation is to protect the overall HSPA performance from few packet losses occurred at the transport network for the best effort traffic. Packet losses occur either due to unreliable wireless channels or congestion caused by the transport network and they can be recovered by the RLC protocol without affecting the upper layer protocols such as TCP. The summary of the implemented protocol functionalities within the HSPA simulator is described in the next sections.
3.6.2
RLC AM mode implementation in HSPA simulator
To achieve the objectives described above, some of the RLC AM functionalities specified in 3GPP [22] are implemented in the HSPA simulator. The implemented functionalities are marked with () whereas others are marked with () in the list given below. The RLC AM mode is applied to each user flow separately between RNC and UE.
Segmentation and reassembly: function performs segmentation and reassembly of variable-length upper layer PDUs into/from RLC PDUs. Concatenation: if the contents of an RLC SDU cannot be carried by one
RLC PDU, the first segment of the next RLC SDU may be put into the RLC PDU in concatenation with the last segment of the previous RLC SDU. A PDU containing the last segment of a RLC SDU will be padded. The effectiveness of this functionality is not significant for a network which uses a large packet sizes such as best effort IP packets. Therefore, this function is not necessary for the HSPA model.
Padding: If concatenation is not applicable and the remaining data to be transmitted does not fill an entire RLC PDU of given size, the remainder of the data field shall be filled with padding bits.
Transfer of user data: this function is used for conveyance of data between users of RLC services.
Error correction: function provides error correction by retransmission (e.g. Selective Repeat) in acknowledged data transfer mode.
In-sequence delivery of upper layer PDUs and duplicate detection: the data will be sent between peer RLC entities in a sequential order. The function detects any arrival of duplicated RLC PDUs and ensures in- sequence delivery to the upper layer.
Flow control: this function allows an RLC receiver to control the rate at which the peer RLC transmitting entity may send.
Sequence number check: this function is used in acknowledged mode. It guarantees the integrity of reassembled PDUs and provides a mechanism for the detection of corrupted RLC SDUs through checking the sequence number in RLC PDUs when they are reassembled into a RLC SDU. A corrupted RLC SDU will be discarded.
Protocol error detection and recovery: this function detects and recovers errors in the AM operation of the RLC protocol.
Ciphering: this function prevents unauthorised acquisition of data. Ciphering is performed in the RLC layer when the non-transparent RLC mode is used. In this simulation, ciphering is not used because the model is implemented in order to evaluate the RLC performance for the data traffic without considering the security aspects.
According to the RLC functionalities specified above, both RLC transmitter and the receiver entities are implemented in the UE and RNC nodes in the uplink and the downlink.
4 HSDPA flow control and congestion control
The flow control and congestion control are the key TNL features which have been developed to enhance the overall end-to-end performance for the downlink of the HSPA network. The motivation and theoretical aspects of flow control are described in this section. A new adaptive credit-based flow control algorithm is introduced for HSDPA. This enhanced flow control scheme has been implemented, tested and validated using the HSPA simulator. Comprehensive simulation results of the flow control are presented here. Since the flow control scheme itself cannot avoid congestion in the transport network, a new congestion control concept has been introduced combined with the adaptive flow control scheme. First the concept of congestion control is described in detail and next different variants of congestion control algorithms are presented which have been implemented, tested and validated within the HSPA simulator for the downlink. Finally a comprehensive simulation investigation and analysis has been performed in order to analyse the end user performance. These two congestion and flow control algorithms address and overcome the two separate issues in the transport network by significantly improving the end- to-end performance. Further these schemes optimise the usage of transport network resources for the HSDPA network while providing aforementioned achievements.
HSUPA also uses the same congestion control scheme as HSDPA. There is no requirement of implementing a flow control scheme since the Node-B scheduler has full control over the transport and radio resources, and further there is no bottleneck behind the RNC in the uplink direction. However a congestion control scheme is required to avoid congestion at the transport level and to optimise the radio resource utilisation. The congestion control input is taken into account in the implementation of the uplink scheduler as described in chapter 3.4.