5GT federation simulations

This repository is oriented to perform simulations for federation interactions in 5G networks.

Clone with submodule

This repository uses the vnfs-mapping submodule to generate the network service graphs. To clone with the submodule run:

git clone --recursive https://github.com/MartinPJorge/5gt-federation.git

Dependencies

Following python packages must be installed for code execution:

pip install networkx==1.9.1
pip install AgglomCluster

NS jukebox

The NS jukebox tool utils/ns_jukebox.py randomly generates network service graphs based on a set of characteristics specified within a configuration file as the one in utils/config/nsGenConf.json.

These are the parameters available in the configuration file:

• splits: number of splits within the NS graph. A split is the point in the network service graph where a VNF1 receives traffic only from VNF0, but then forwards traffic to VNF2...VNFn;
• splitWidth: number of VNFs after the VNF split, if VNF1 forwards traffic towards VNF2, VNF3, VNF4; the split width is equal to 3;
• branches: maximum number of parallel flows within the NS graph;
• vnfs: maximum number of VNFs within the NS graph;
• linkTh: (min,max) values to generate the virtual links characteristics below. If linkTh["traffic"]["min"]=2 and linkTh["traffic"]["min"]=5, then the algorithm can generate a bandwidth demand between 2 and 5.
  • linkTh["traffic"]: (min,max) interval for virtual link bandwidth;
  • linkTh["delay"]: (min,max) interval for virtual link bandwidth;
• vnfTh: (min,max) values to generate the VNF characteristics below.
  • vnfTh["processing-time"]: VNF processing time required by a VNF;
  • vnfTh["processing-time"]["cpu"]: VNF required CPU;
  • vnfTh["processing-time"]["memory"]: VNF required memory;
  • vnfTh["processing-time"]["storage"]: VNF required disk;

Execution example

Following example generates 2 NS graphs based on the example configuration file, and stores the virtual links and VNFs CSVs under /tmp:

$ python ns_jukebox.py config/nsGenConf.json 2 /tmp
Generating 0-th NS for config config/nsGenConf.json
  VNF CSV at: /tmp/vls-0.csv
  VL CSV at: /tmp/vnfs-0.csv
Generating 1-th NS for config config/nsGenConf.json
  VNF CSV at: /tmp/vls-1.csv
  VL CSV at: /tmp/vnfs-1.csv

Output CSV examples

For each generated network service, two files are created. One for the virtual links:

delay,traffic,prob,idVNFa,idVNFb
5,502,0.7904173707373455,1,2
2,105,0.20958262926265447,1,3
2,238,1,2,4
3,39,1,3,4
4,712,1,4,5
1,882,1,5,6
2,528,1,6,7
2,767,1,7,8
5,971,1,8,9
4,880,1,9,10

and another for the VNFs:

requirements,processing_time,id,memory,vnf_name,disk,cpu,idVNF
{},3,1,561,v_gen_1_4_1553105587.72,4753,23,1
{},3,2,842,v_gen_2_3_1553105587.72,9639,3,2
{},2,3,997,v_gen_3_2_1553105587.72,9654,9,3
{},19,4,864,v_gen_4_16_1553105587.73,4212,24,4
{},1,5,821,v_gen_5_5_1553105587.73,3482,14,5
{},19,6,219,v_gen_6_1_1553105587.73,8530,9,6
{},15,7,631,v_gen_7_3_1553105587.73,7358,9,7
{},7,8,64,v_gen_8_19_1553105587.73,6339,18,8
{},16,9,91,v_gen_9_3_1553105587.73,7679,14,9
{},8,10,516,v_gen_10_13_1553105587.73,5315,13,10

NS splitter

This script performs the clustering of a network service graph given the virtual links and VNFs CSV files.

It is executed running:

python ns_splitter.py /tmp/vnfs-1.csv /tmp/vls-1.csv /tmp

where first and second arguments are the VNFs and VLs CSV files generated with ns_jukebox.py, respectively. And last parameter is the directory where the clusters are stored following naming convention ns-NS_NUM-clusters-NUM_CLUSTERS-clusterCLUSTER_NUM-(vls|vnfs).csv. That is, the script tries to create from 1 up to len(VNF) clusters, then it generates CSV files for the VNFs and VLs within each cluster.

=== 2 CLUSTERS ===
                              | ns-1-clusters-2-cluster0-vnfs.csv
                 __cluster-0__| ns-1-clusters-2-cluster0-vls.csv
                /
vnfs-1.csv |___/
vls-1.csv  |   \              | ns-1-clusters-2-cluster1-vnfs.csv
                \__cluster-1__| ns-1-clusters-2-cluster1-vls.csv

=== 3 CLUSTERS ===
                            
                 __cluster-0__ ...
                /
vnfs-1.csv |___/___cluster-1__ ...
vls-1.csv  |   \            
                \__cluster-2__ ...

[...]

Acknowledgements

5G-TRANSFORMER Project under Grant 761536. Parts of this paper have been published in the Proceedings of the IEEE BMSB 2018, Valencia, Spain.