Modelling soft handover

Soft Handover (SHO) has some negative influences on the network performance. Especially, the influence of SHO on the network performance and the individual end user performance are significant at overload situations. SHO leads to additional load inside the transport network. Therefore it causes blocking of certain parts of the BW for these additional overheads. The transport congestion control algorithm also has to handle such HO situations effectively. Therefore within the focus of the TNL feature development in the network simulator, it is important to consider the effect of the SHO load on the transport network. This leads to modelling of SHO effects in the uplink simulator.

For uplink transmission, the E-DCH scheduler has to control the offered load to the transport network. In a congestion situation, the scheduler should reduce the offered load to the transport network for some greedy connection in the uplink. These actions are performed combined with the transport network congestion control algorithm. Since SHO is causing additional load over the transport network, the E-DCH scheduler has to manage the uplink transport resources among user connections during the soft handovers. However SHO is not critical for the downlink since the Node-B is centrally managing the resources among connections and the flow control scheme along with the downlink transport congestion control scheme decides the offered load to the transport network. An overload situation is completely controlled by the DL congestion control schemes and further, in this case, the main buffering for the connections is located at the RNC entity. Therefore, for transport analysis, DL Soft handover does not play a significant role on the network and end user performance. Within the discussion of this section mainly the effect of the soft handover for uplink transmission is modelled. For uplink transmission, two different SHO cases can be distinguished: the first case represents the soft handover that some of the Serving Radio Link Set (S-RLS) UEs are experiencing, in which those UEs will have their packets carried over the N-SRLS Iub link to the RNC. The second case

represents the UEs from the neighbouring cells that are in soft handover with the current cell which belongs to the considered Node-B. Those UEs are referred to as N-SRLS UEs, and some of their packets have to be carried by its Iub link.

Add RLC headers RLC segmentation process AAL2 ATM Application TCP/UDP IP Air Interface PerUE RLC buffers MAC-es buffers N*MAC-d MAC-e Txn and Retxn buffers

High Loss probability (ex. 40%-60%) X% traffic NRLS UEs AAL2 ATM FP header processing MAC-es buffers MAC-es header processing PerUE RLC buffers Application TCP/UDP IP Iub Link NSRLS UEs FP _delayIub SHO Ues Transmitting via the NSRLS

Emulating the delay between the other Node-B and RNC MAC-es MAC-d RLC MAC-e MAC-e EDCH-FP EDCH-FP MAC-es MAC-d RLC Node-B UE RNC

Figure 3-11: architectural overview of UL UE modelling with SHO functionality

Figure 3-11 shows the architectural overview of the UL UE modelling with SHO functionality. Further, it is shown that the SHO UE data is transmitted via neighbouring Iub link and the N-SRLS UE data is transmitted over the

considered Iub link. The first case (S-RLS UEs in SHO) is modelled in the following way:

a) The number of SHO UEs is chosen as a percentage of the total number of the HSUPA UEs.

b) The SHO UEs are considered to be in handover during the whole simulation time. This is done in order to emulate the additional load on the transport network.

When SHO UEs are transmitting a MACe PDU over the air interface, a copy of the packet is also transmitted via the Iub of the neighbouring cells. Since these neighbouring cells are not modelled in this simulator however, they are modelled in following way to counter the impact of the load offered.

A direct link was created in the model that links the HSUPA layer of the Node-B with the corresponding layer of the RNC. This link is used to emulate the effects of the N-SRLS Iub links, where the packets are transmitted over this link directly to the RNC by emulating a certain packet error rate (for example 40% – 60%) and a certain delay variation (for example, 10 ms – 40 ms). Therefore SHO UEs get the duplicate arrivals at RNC and they are filtered at the MAC-es layer in RNC.

The N-SRLS UEs that are in soft handover with the cell in consideration are modelled in the simulator as well. Since the N-SRLS UEs are not the main concern of this analysis, and also the per-UE evaluation is not intended for them there is no need of modelling the full protocol architecture for those N- SRLS UEs to generate their traffic, instead a simplified approach is used to generate the traffic for those UEs according to the following guide lines. The N-SRLS UEs are needed to block some of the Iub link capacity with their traffic to have the realistic scenarios for the simulations, so those UEs should only generate MACe PDUs to be carried over the Iub (after being encapsulated as FP PDUs) and then discarded on the RNC when they are received.