Network Slicing with NetSim

Overview:

Network slicing enables the creation of multiple virtual networks, or slices, on a shared physical infrastructure. Operators need to efficiently allocate RAN resources to meet SLAs across different service categories within each slice.

  • SLA Assurance: SLAs must be met despite challenges such as user mobility, varying channel quality, and network load fluctuations.
  • Resource Allocation Complexity: Ensuring QoS through dynamic resource reservation is a hard problem. RAN is the bottleneck due to its reliance on the most expensive resource - spectrum.

NetSim Feature Set:

  • RAN Slicing: Define multiple slices at the network level.
  • Configuration: Set up S-NSSAI, which includes SST (Slice/Service Type) and SD (Slice Differentiator). Associate UEs with slices.
  • Resource Sharing Options:

    • Static Allocation:

      Allocate a fixed percentage of resources.

    • Dynamic Allocation:

      A proprietary online machine learning algorithm allocates resources at run time, to meet SLAs "statistically".
  • Telemetry: gNB and UE level radio measurements, Per UE and per gNB MAC throughput, Per gNB PRB utilization, Per slice PRB utilization and more. NetSim gives plots and detailed CSV logs.

Benefits of NetSim Simulation:

  • Simulate scenarios with network slices, and check if the network meets the SLAs.
  • Do what-if analysis to evaluate different network configurations and their effect on performance.
  • Test and optimize your slicing algorithms in a controlled environment before real-world deployment.

Customization:

NetSim includes source C code. Modify it to develop and test your own slicing algorithms.


RAN network slicing in NetSim: slices configured through the GUI, with the RAN Intelligent Controller allocating physical resource blocks across slices.

RAN Network Slicing in NetSim: configure the slices through the GUI. The RAN Intelligent Controller (RIC), with AI/ML capabilities, allocates the radio resources (PRBs) across slices. It interacts with the PHY, MAC and RLC layers. NetSim provides the source code, so you can implement your slicing algorithms in the RIC.

Example Use Case

A. Baseline scenario with 3 UEs at different distances, PFS scheduling (No slicing)

The scenario has 3 UEs at different distances, and each UE has a time-varying radio channel. The proportional fair algorithm schedules them. These are the results:

NetSim scenario: gNB_9 serving UE_10 at 1500 m, UE_11 at 2000 m and UE_12 at 2500 m, backhauled through core switches, UPF, AMF and SMF to Server_8, with downlink constant-rate UDP flows to each UE
System Parameters
CA Configuration n78
Channel Bandwidth 100 MHz
Pathloss Model 3GPPTR38.901-7.4.1
Outdoor Scenario Urban Macro
Fading Rayleigh
Scheduling Type Proportional Fair
EWMA Learning Rate 0.005
NetSim plot of EWMA MAC throughput against time for UE 10, UE 11 and UE 12 under proportional fair scheduling, showing the three UEs separated at roughly 70, 45 and 25 Mbps
UE ID Avg. MAC Throughput (Mbps)
10 70.95
11 43.59
12 24.50

B. Network slicing to provide a 30 Mbps GBR to UE12

Case (a) shows that UE12 gets only 24.50 Mbps. To meet the 30 Mbps target, UE12 needs more PRBs. We create an eMBB slice and associate UE 12 to it. We see NetSim results from two slicing options: static resource allocation and dynamic resource allocation.

Static slicing: EWMA MAC throughput against time for UE 10, UE 11 and UE 12, with UE 10 near 45 Mbps and UE 11 and UE 12 converged near 25 Mbps
Dynamic slicing: EWMA MAC throughput against time for the three UEs, with an inset plot showing the algorithm's index bias rising while throughput is below the 30 Mbps target and falling once it is above
UE ID Throughput without
slicing (Mbps)		 Throughput with Static
Resource Allocation (Mbps)		 Throughput with Dynamic
Resource Allocation (Mbps)
10 70.95 35.58 62.68
11 43.59 21.15 38.73
12 24.50 30.00 (Thput. guarantee 30 Mbps) 30.02 (Thput. guarantee 30 Mbps)

Sum

139.04

86.73

131.35

We see that dynamic resource allocation achieves a 51.4% higher throughput while meeting the throughput guarantee for UE12.

C. Configuring multiple slices with dynamic resource allocation in NetSim

We consider the same scenario as in (a) and create two eMBB slices. We associate UE12 with one slice and UE11 with the other slice. Then we configure a GBR of 30 Mbps for UE12 and 45 Mbps for UE11.

The same NetSim scenario configured with three slices: slice 1 best-effort for UE 10, slice 2 eMBB for UE 11 and slice 3 eMBB for UE 12, posing the question of whether 30 Mbps to UE 12 and 45 Mbps to UE 11 can be guaranteed together
EWMA MAC throughput against time with two eMBB slices active, showing multiple throughput guarantees met simultaneously, alongside an inset index-bias plot for UE 11 and UE 12
UE ID Slices Throughput (Mbps)
10 Slice 1, BE 52.44
11 Slice 2 (Thput. guarantee 45 Mbps) 45.01
12 Slice 3 (Thput. guarantee 30 Mbps) 30.02

Sum throughput

127.31

NetSim's in-built dynamic resource allocation algorithm distributes resources between the slices, and both slices meet their guaranteed bit rates.