SliceSim: A Simulation Suite for Network Slicing in 5G Networks

Abdurrahman Dilmaç (abdurrahman.d at icloud dot com)
Muhammed Emin Güre (memingure at gmail dot com)
Project Advisor: Prof. Tuna Tuğcu

Introduction

5G widely defines network slicing concept which aims to provide different and separate dedicated logical networks that can be customized to respective services. All slices under a cloud infrastructure are put together with their different requirements, e.g. bandwidth, latency

The purpose of this project is to provide a simulation suite for a network consisting of base stations and clients that possible scenarios of 5G can fit into and make analysis of different concepts easier.

Approach

Discrete event simulation
Using Python 3.7, Simpy, Matplotlib, KDTree
YAML for reading input configurations
Asynchronous programming
Definitions:
- Clients: Simulation consumers. Generates consume requests by given distribution parameters.
- Slices of Base Stations: Simulation resources.

Input

Settings

settings:
  simulation_time: 100 # in seconds
  num_clients: 100
  limit_closest_base_stations: 5 # how many base stations stored in a client instance
  statistics_params:
    warmup_ratio: 0.05 # statistic collection will start from this point
    cooldown_ratio: 0.05 # statistic collection will end at this point
    # Statistic collection will be in this area of the coordinate system
    x:
      min: 0
      max: 1980
    y:
      min: 0
      max: 1980
  logging: False # saving logs to a file
  log_file: output.txt # name of the log file
  plotting_params:
    plotting: True # plot the statistics after execution
    plot_save: True # save plot as image
    plot_show: False # show plot after execution
    plot_file: output.png # name of the plot image
    plot_file_dpi: 1000 # dots per inch for plot image
    scatter_size: 15

Slices

slices:
  slice_name: # name of the slice
    delay_tolerance: 10
    qos_class: 5
    bandwidth_guaranteed: 0 # in bps
    bandwidth_max: 100000000 # in bps
    client_weight: 0.39 # [0,1] - ratio of the clients subscribed to this slice in the system. All weights for slices must be 1 in total
    threshold: 0 # for dynamic slicing (future work)
    # defines the bit usage pattern for client subscribed to this slice
    usage_pattern:
      distribution: randint # distribution name
      params: # distribution parameters
        - 4000000 # min value for this example
        - 800000000 # max value for this example
  slice_name_2:
  	...

Base Stations

base_stations:
  - x: 182 # in meters
    y: 1414 # in meters
    capacity_bandwidth: 20000000000 # in bps
    coverage: 224 # in meters
    # ratios of the slices in this specific base station
    # must be 1 in total for each base station
    ratios:
      slice_name: 0.20 # [0,1]
      slice_name_2: 0.59 # [0,1]
      ...
  - ...

Mobility Patterns

mobility_patterns:
  mobility_pattern_1:
    distribution: normal # distribution name
    params: # distribution parameters
      - 0 # mean value for this example
      - 7 # standard deviation name for this example
    client_weight: 0.10 # [0,1] - ratio of the clients assigned to this pattern in the system. All weights must be 1 in total
  mobility_pattern_2:
  	...

Client population

clients:
  location: # populates the area with the given distributions
    x:
      distribution: randint
      params:
        - 0
        - 1980
    y:
      distribution: randint
      params:
        - 0
        - 1980
  usage_frequency: # defines the usage generation intervals of clients
    distribution: randint
    params:
      - 0
      - 100000
    divide_scale: 1000000 # scaling factor

Usage

python -m slicesim <input-file.yml>

Example Output

Conclusion

Increasing number of clients increases used bandwidth, and yet the simulation showed that block ratio also elevates for this specific configurations.

This simulation tool can be used for such scenarios as well:

Testing the effect of different dynamic slicing algorithms on block and handover ratios.
Analyzing various mobility patterns of clients using different statistical distributions.
Observing the effect of usage frequency of clients and the effect of clients those are distributed unequally.
Various Proof of Concepts like common base stations for multiple service providers.

Future Work

Customizable shapes for base station coverages
Improvements of the software performance
Dynamic slicing mechanism
Generation of more test configurations
Video output of a running simulation

References

5GPPP Architecture Working Group. View on 5G Architecture. Version 2.0, December 2017
CellMapper - https://www.cellmapper.net (10.05.2019)
FatihMunicipalityGeographicInformationSystem- https://gis.fatih.bel.tr/webgis (13.05.2019)
https://venturebeat.com/2018/12/12/decoding-5g-a-cheat- sheet-for-next-gen-cellular-concepts-and-jargon/ (17.03.2019)

task-master98/slicesim