QoS Scheduler Evaluation

For more information about the QoS scheduler, visit 5G-LENA Doxygen

Simulation Scenario

---

1. 3 UEs with a single flow

   There are 3 UEs attached to a gNB. Two UEs have non-GBR traffic with a 5QI of 80, and another UE has DC GBR traffic with a 5QI of 87.

2. 2UEs with multi-flow

   There are 2 UEs attached to a gNB. One has non-GBR traffic with a 5QI of 80, and the other has two flows: non-GBR traffic with a 5QI of 80 and DC GBR traffic with a 5QI of 87.

Simulation Environment 1

---

bandwidth	10MHz (saturation)
numerology	0
traffic type	2 (5QI 80 and 87)
priorityTrafficScenario	1 (medium load)
packet size	5QI 80 = 3000 bytes / 5QI 87 = 1252 bytes
lambda	1000

3 UEs with a single flow

# Tx Packets 600

Flow	UE1 (non-GBR. 5QI=80)	UE2 (non-GBR. 5QI=80)	UE3 (DC GBR. 5QI=87)
Throughput	18.046 Mbps	18.087 Mbps	10.171 Mbps
Mean delay	82.110 ms	82.269 ms	4.066 ms
Mean jitter	0.364 ms	0.318 ms	0.015 ms
# Rx Packets	448	448	596

• Mean flow throughput: 15.435 Mbps
• Mean flow delay: 56.148 ms

2 UEs with multi-flow (QoS-RR)

# Tx Packets 600

Flow	UE1 (non-GBR. 5QI=80)	UE2-1 (non-GBR. 5QI=80)	UE2-2 (DC GBR. 5QI=87)
Throughput	11.587 Mbps	19.056 Mbps	10.171 Mbps
Mean delay	161.440 ms	69.313 ms	4.083 ms
Mean jitter	1.046 ms	0.252 ms	0.013 ms
# Rx Packets	287	472	596

• Mean flow throughput: 13.605 Mbps
• Mean flow delay: 78.2791 ms

2 UEs with multi-flow (QoS-QoS)

# Tx Packets 600

Flow	UE1 (non-GBR. 5QI=80)	UE2-1 (non-GBR. 5QI=80)	UE2-2 (DC GBR. 5QI=87)
Throughput	11.587 Mbps	16.916 Mbps	10.171 Mbps
Mean delay	161.374 ms	96.246 ms	4.077 ms
Mean jitter	1.046 ms	0.407 ms	0.013 ms
# Rx Packets	287	419	596

• Mean flow throughput: 12.891 Mbps
• Mean flow delay: 87.232 ms

Simulation Environment 2

---

bandwidth	10MHz (saturation)
numerology	0
traffic type	2 (5QI 80 and 87)
priorityTrafficScenario	0 (saturation)
packet size	3000 bytes
lambda	1000

3 UEs with a single flow

# Tx Packets 600

Flow	UE1 (non-GBR. 5QI=80)	UE2 (non-GBR. 5QI=80)	UE3 (DC GBR. 5QI=87)
Throughput	0.444 Mbps	0.444 Mbps	23.820 Mbps
Mean delay	323.697 ms	324.645 ms	10.026 ms
Mean jitter	48.272 ms	48.188 ms	0.005 ms
# Rx Packets	11	11	590

• Mean flow throughput: 8.236 Mbps
• Mean flow delay: 219.456 ms

2 UEs with multi-flow (QoS-RR)

# Tx Packets 600

Flow	UE1 (non-GBR. 5QI=80)	UE2-1 (non-GBR. 5QI=80)	UE2-2 (DC GBR. 5QI=87)
Throughput	0.444 Mbps	23.820 Mbps	23.820 Mbps
Mean delay	324.242 ms	10.026 ms	10.026 ms
Mean jitter	48.090 ms	0.003 ms	0.003 ms
# Rx Packets	11	590	590

• Mean flow throughput: 16.028 Mbps
• Mean flow delay: 114.765 ms

2 UEs with multi-flow (QoS-QoS)

# Tx Packets 600

Flow	UE1 (non-GBR. 5QI=80)	UE2-1 (non-GBR. 5QI=80)	UE2-2 (DC GBR. 5QI=87)
Throughput	0.444 Mbps	22.124587 Mbps	23.820 Mbps
Mean delay	324.242 ms	31.475513 ms	10.026 ms
Mean jitter	48.090 ms	0.078206 ms	0.003 ms
## Rx Packets	11	548	590

• Mean flow throughput: 15.462 Mbps
• Mean flow delay: 121.915 ms

Conclusion

---

• When there is only a single flow in each UE, I observe that all non-GBR traffic experiences more severe delays compared to DC GBR traffic.
• As I analyzed the simulation results from the QoS scheduler paper, non-GBR traffic benefits when the UE with non-GBR traffic also has DC GBR traffic in a multi-flow configuration.

Questions

---

1. Why does the DC GBR flow not gain as much as the non-GBR flow of UE2 loses when switching the scheduler from QoS-RR to QoS-QoS?
2. In Scenario 2 of Env.2, the non-GBR flow of UE2 is benefiting from the weight of the DC-GBR flow in scheduling. Therefore, it is expected that in Scenario 1, where there is no such benefit, UE1 and UE2 would share the delay equally and experience lower delays compared to the delay of UE1 in Scenario 2. However, both show similarly high delays of approximately 300ms. Why is this the case?