In Chapter 1 the benefits and disadvantages of soft handover in UMTS FDD mode were discussed. Now two KPIs are defined that measure the percentage of disadvantages: primary and secondary RLC traffic. As will be demonstrated these KPIs can only be calculated at cell level and a mandatory prerequisite is that during an active connection the performance measurement software always knows which cells belong to the active set of this connection.
Primary traffic is the traffic that is necessary to transmit data between the UE and the network and vice versa. If the UE is in a softer or soft handover situation identical RLC transport blocks are sent/received via each radio link belonging to the active set of the UE.
One radio link is always related to a single cell. If there are e.g. two radio links related to two cells in an active set, double the amount of uplink and downlink RLC transport blocks is sent on the radio interface to benefit from the advantages of the soft handover scenario. One-half of this doubled traffic volume is called primary traffic, the other half is called the secondary traffic. It is important to understand that calculation of primary and secondary traffic follows straightforward statistical rules. There is no preference of a single radio link or a single cell in this formula.
Note: a logical failure that is often heard when discussing primary and secondary traffic is the assumption that all primary traffic is sent via the best cell of connection while all other links in the active set are only used to transmit secondary traffic.
The truth is that errors in radio transmission can occur on each link belonging to an active set. Hence, for the transmission of a certain RLC transport block it is not known which cell is the best cell to transmit this block error-free. If a cell offered such a high quality that transmission errors could be excluded the ongoing connection would not be in a soft hand- over situation. Following this statistically, primary and secondary traffic are seen as distri- buted to all radio links of the active set. The basic rule that is illustrated in Figure 2.24 says:
1
2
3
RNC
RNC UE
UE
UE
Cell 1
Cell 1
Cell 1
Cell 2
Cell 3 Cell 2
100 % of transmitted data monitored in Cell 1
100 % of transmitted data monitored in Cell 1
100 % of transmitted data monitored in Cell 1 100 % of transmitted data monitored in Cell 2
100 % of transmitted data monitored in Cell 2
100 % of transmitted data monitored in Cell 3
SRNC
SRNC
RNC
SRNC All traffic is primary traffic = 100 % Primary and Secondary Traffic
Primary Traffic per Cell Secondary Traffic
per Cell
=
= 100% – 50% = 50%
100% =50%
2
Primary Traffic per Cell Secondary Traffic
per Cell
=
= 100% – 33.33% = 66.67%
100% = 33.33%
3
Figure 2.24 Primary and secondary traffic in three different cases
the more radio links a UE has in its active set the more secondary traffic and the less primary traffic is measured for this single UE connection.
To calculate primary and secondary traffic it is not necessary to measure the data volume or throughput of a connection. Looking at the different situations shown in Figure 2.24 it becomes clear that primary and secondary traffic as a percentage value can be calculated based on the average active set size of a single connection as shown in Equations (2.11) and (2.12). To aggregate primary and secondary traffic at a different aggregation level than the call (e.g. per cell or per SRNC) requires the average active set size of all connections that have been active using these network resources during a defined time period.
Primary Trafficẳ 1
Average Active Set Size of connection100% ð2:11ị Secondary Trafficẳ100%Primary Trafficð%ị ð2:12ị It can be assumed that network operators would like to express how much radio trans- mission capacity is provided to transmit primary and secondary traffic, because minimising the secondary traffic component indicates the increase of the total capacity of the network.
For this reason percentage values are combined with uplink and downlink RLC throughput measured on lub/Iur physical transport bearers (AAL2 SVC) used to transport data streams of dedicated traffic channels (other channels are not involved when the UE is in soft hand- over). Although this calculation does not take into account the throughput caused by the FP header and trailer information bits it allows a good estimation of how much bandwidth is necessary to provide transport resources for secondary traffic in soft handover situations. For softer handover scenarios (cells involved in soft handover are located in the same Node B) this calculation does not apply, because usually Node B performs the combining of radio links and hence only a single AAL2 SVC is used on the Iub interface, but data is transmitted on two radio links in parallel.
Now the target is to find out how many additional resources are necessary to transmit secondary traffic. On wired interfaces, Iub and Iur, this is the bandwidth of electrical or opti- cal cables, on the radio interface, Uu, bandwidth is correlated with the channelisation codes used in downlink. The number of available downlink channelisation codes limits the number of possible connections and services per cell.
Code utilisation per cell should rather be analysed by using a separate KPI, but transport overhead due to secondary traffic on Iub and Iur can be computed if the following preposi- tions are fulfilled:
Performance measurement software must be able to distinguish between softer handover when diversity splitting/combining is done in Node B and soft handover situations when diversity splitting/combining is done in RNC.
The overall measured transport channel throughput based on the size of transport blocks must be associated to each physical transport bearer of a single connection.
Instead of the average active set size the average size of the used transport bearer set of connection must be known, which requires a special application that works in a similar way as the active set size tracking, but instead of the set up and deletion of
radio links per connection this other application must track the set up and deletion of VPI/VCI/CID per connection.
Secondary traffic on lub/Iur physical transport resources could then be computed as shown in Equation (2.13):
1 1
Average Size of TransportBearerSet
Transport Channel Throughput ẵkbps ð2:13ị
Whether the separation of uplink and downlink measurements is necessary may be decided for each particular case. In an ideal case the throughput on physical transport bearers can be measured at frame protocol level (including FP header and trailer information bits) and a topology detection tool provides additional information about which VPI/VCI/CID belong to which physical wire or fibre between network elements. Measurement results can be correlated with fixed costs of transport network resources, which especially in the case of leased lines can be expressed in costs per kbps bandwidth. The optimisation of secondary traffic can help to minimise these fixed costs. However, it must be kept in mind that the minimisation of secondary traffic is only possible as far as overall radio transmission is guaranteed and smooth soft handover procedures are guaranteed. Thus the optimisation of primary/secondary traffic is one of the most complex and most difficult procedures.
Another idea that is sometimes seen in requirement documents is the correlation of primary/secondary traffic percentage with transmitted data volume at the RLC SDU level, which is the level of user-perceived throughput. The objective behind such formulas is to estimate for how much traffic transmitted on UTRAN interfaces the user will be charged and how much ‘payload’ needs to be transmitted free of charge due to soft handover. Depending on how the fixed costs of network operation are distributed (e.g. per cell or per wire/fibre) the data volume of SDUs transmitted on uplink and downlink need to be measured and measurement values need to be distributed in the same way as cost (also per cell or per wire/fibre). If measurement results are computed at cell level, active set size tracking is necessary; at transport bearer level the same prerequisites apply as described for transport channel throughput.
The data volume of charged payload at cell level can be computed as shown in Equa- tion (2.14):
DataVolume of ULþDL RLC SDUs per cell
Average Active Set Size per cell ð2:14ị
The ‘per cell’ aggregation level in this formula is realised by calculating an average (active set size) or total value (data volume) of all connections that use dedicated radio links provided by this cell. There must be a high measurement tolerance taken into account due to fact that large SDUs are segmented into a huge number of transport blocks and the cell used for the transmission of these transport blocks can be changed with every transport block set that is sent or received.