NetSim Support Portal

Use Case: Impact of numerology on throughput and latency of a 5G NR scenario having 1 gNB connected to 25 phones, 6 sensors, and 3 cameras, with multiple TCP and UDP flows

Network Scenario: To model a real-world scenario, we base our simulation on the setup shown in Figure 1. At a high level, the link between the gNB and the L2_Switches that represents the Core Network (CN) is made with a point-to-point 10 Gb/s link, without propagation delay. The Radio Area Network (RAN), is served by 1 gNB, in which different UEs share the connectivity. We have 25 smartphones, 6 sensors, 3 IP cameras. The bandwidth 100MHz and Round Robin MAC Scheduler. The position of the devices in the reference scenario depicted in Figure 1 is quasi-random. 

Fig 1: Network scenario with 25 smartphones, 6 sensors, and 3 cameras communicating with respective cloud servers

 In terms of application data traffic, the camera (video) and sensor nodes have one UDP flow each, that goes in the UL towards a remote node on the Internet. These flows are fixed-rate flows: we have a continuous transmission of 5 Mb/s for the video nodes, to simulate a 720p24 HD video, and the sensors transmit a payload of 500 bytes each 2.5 ms, that gives a rate of 1.6 Mb/s. For smartphones, we use TCP as the transmission protocol. These connect to database servers. Each phone has to download a 25 MB file and to upload one file of 1.5 MB. These flows start at different times: the upload starts at a random time between the 25^th and the 75^th simulation seconds, while each download starts at a random time between the 1.5^th and the 95^th simulation seconds.

Table 1: Various parameters of the Traffic flow models for all the devices

The numerology μ can take values from 0 to 3 and specifies an SCS of 15 x 2^μ kHz and a slot length of 1/(2^μ) ms. FR1 supports μ = 0, 1, and 2, while FR2 supports μ = 2 and 3.   We study the impact of different numerologies, and how they affect the end-to-end performance. The metrics measured and analyzed are a) Throughput of TCP uploads & downloads, and b) Latency of the UDP uploads

Results and analysis

Fig 2 Camera Uplink, and Sensor Uplink average throughput vs. Numerology (µ)

Fig 3 Smartphone Uplink, and Smartphone Downlink average throughput vs. Numerology (µ)

 Fig 4 Camera Uplink, and Sensor Uplink Latency vs. Numerology

For UDP applications the μ does not impact the throughput. However, higher μ leads to an obviously lower delay. The variation of delay vs. μ  is as follows:

Fig 5 Camera Uplink, and Sensor Uplink average throughput vs. Numerology (µ)

Fig 6 Smartphone Uplink, and Smartphone Downlink average throughput vs. Numerology (µ)

Fig 7 Camera Uplink, and Sensor Uplink Latency vs. Numerology

For UDP applications the μ does not impact the throughput. However, higher μ leads to an obviously lower delay. The variation of delay vs. μ  is as follows:

The TCP throughput is inversely proportional to round trip time. Therefore, for applications running over TCP the throughput increases with higher numerology. This is because higher μ leads to reduced round-trip (end-to-end) times. 

References

1. Natale Patriciello, Sandra Lagen, Lorenza Giupponi, Biljana Bojovic.”5G New Radio Numerologies and their Impact on the End-To-End Latency” in 2018 IEEE 23rd International Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD)

Appendix 1: Settings done in NetSim for modeling this network

Application Properties set in NetSim (Based on Table 1)

Appendix 2: Detailed Results

The network Configuration files associated with all the three cases considered for obtaining the results mentioned above are attached to this article. The Configuration.netsim file associated with each case can be imported into NetSim using the Import Experiment Option in the Open Simulation menu of NetSim Home Screen.