MAXCI Algorithm

The retransmission processes unconditionally have higher priorities than the initial transmission queues. The retransmission processes are sorted in first-in first-out (FIFO) mode. The initial transmission queues are sorted in the CQI order. A higher CQI means a higher data priority.

The MAXCI algorithm aims to maximize the system capacity but cannot ensure user fairness and differentiated services. RR Algorithm

The retransmission processes unconditionally have higher priorities than the initial transmission queues. The retransmission processes are sorted in FIFO mode. The initial transmission queues are sorted in the order of the waiting time in the MAC-hs queue. A longer waiting time means a higher data priority.

The RR algorithm aims to ensure user fairness but cannot provide differentiated services. Not considering the CQI reported by the UE leads to lower system capacity.

PF Algorithm

The retransmission processes unconditionally have higher priorities than the initial transmission queues. The retransmission processes are sorted in FIFO mode. The initial transmission queues are sorted in the order of R/r. Here, R represents the throughput corresponding to the CQI reported by the UE, and r represents the throughput achieved by the UE. A greater R/r value means a higher data priority.

The PF algorithm aims to make a tradeoff between system capacity and user fairness. It provides the user with an average throughput that is proportional to the actual channel quality. The system capacity provided by PF is between the system capacity provided by RR and that provided by MAXCI.

EPF Algorithm

The EPF algorithm can meet the requirement of the telecom operator for controlling user fairness and differentiated services and provide very high system capacity.

Firstly, the EPF algorithm distinguishes between delay-sensitive data and throughput-sensitive data. The delay-sensitive data unconditionally has a higher priority than the throughput-sensitive data.

The delay-sensitive data includes the following data:

SRB signaling

VoIP and AMR service data whose waiting time approaches the value of the discard timer

The throughput-sensitive data includes the following data:

BE service data

Streaming service data

IMS data

VoIP and AMR service data whose waiting time is far from the value of the discard timer

Secondly, the EPF algorithm distinguishes between retransmission processes and initial transmission queues. The retransmission processes unconditionally have higher priorities than the initial transmission queues.

If two queues or processes have the same absolute priority, the algorithm considers the waiting time, CQI reported by the UE, throughput achieved by the UE, guaranteed bit rate (GBR), scheduling priority indicator (SPI) weight, happy bit rate (HBR), and power consumed in the queue for a certain period, which are described as follows:

For the delay-sensitive data, a longer waiting time means a higher data priority.

For the throughput-sensitive data, a greater R/r value means a higher data priority. Here, R represents the throughput corresponding to the CQI reported by the UE, and r represents the throughput achieved by the UE.

The users not achieving the GBR have higher priorities than those already achieving the GBR.

A higher SPI weight means a higher data priority.

A larger difference between the actual rate and the HBR means a higher data priority.

When the resource limitation switch (RscLmSw) is on, the algorithm gives the lowest priority to a queue whose power consumption exceeds the threshold. RscLmSw is used to prevent the users in a poor-coverage area from consuming too many cell resources and thus avoid a sharp decrease in system capacity. The ratio of the maximum available power of a queue to the total power of the cell depends on the GBR, as shown in Table 5-1.

Table 1-1 Default maximum ratios based on the GBR

GBR (kbit/s) Maximum Ratio

8 10%

16 10%

32 15%

64 15%

128 20%

256 25%

384 30%

The SPI weight is used to control differentiated services for different users. It can be specified on the basis of SPI on the RNC side.

The HBR is used to control user fairness. It defines the highest throughput expected by the user when the resources are limited. By giving more chances to the users without the HBR than those with the HBR, the algorithm improves the overall user satisfaction. The HBR is specified by the parameter HappyBR on the RNC side. The setting can be based on user levels, including gold, silver, and copper.

For details on the parameters related to QoS management, such as the GBR, SPI, SPI weight, and HBR, see section 6.2 "Diff-Serv Management."

The EPF algorithm is optional.

1.1.2 Comparison of Four Algorithms

Table 5-2 lists the factors considered in the four scheduling algorithms.

Table 1-1 Factors considered in the four scheduling algorithms

Factor MAXCI RR PF EPF

Service type No No No Yes

Initial transmission or retransmission Yes Yes Yes Yes

Maximum power No No No Yes

Waiting time No Yes No Yes

CQI Yes No Yes Yes

Actual throughput No No Yes Yes

SPI No No No Yes

GPR No No No Yes

HBR No No No Yes

Table 5-3 lists the effects of the four scheduling algorithms.

Table 1-2 Effects of the four scheduling algorithms

Item MAXCI RR PF EPF

System capacity Highest High Higher Higher

User fairness Not guaranteed Best Guaranteed Guaranteed

Differentiated services Not guaranteed Not guaranteed Not guaranteed Guaranteed

Real-time services Not guaranteed Not guaranteed Not guaranteed Guaranteed