TPG stack on G5K + FIT IoT LAB

Experiment Artifacts

The artifacts repository contains the E2Clab configuration files such as layers_services.yaml, network.yaml, and workflow.yaml

$ cd ~/git/
$ git clone https://gitlab.inria.fr/E2Clab/examples/monitoring-tpg-g5k-iotlab.git
$ cd monitoring-tpg-g5k-iotlab/

Defining the Experimental Environment

Layers & Services Configuration

The monitoring type is tpg with the provider (machine hosting Prometheus and Grafana) on the g5k testbed (paravance cluster). The network is shared (experiment data and monitoring data are on the same network). We use an IPv6 network (FIT IoT LAB devices use IPv6 network). Finally, we added roles: [monitoring] in all services (e.g., MyServer, MyClientA, and MyClientB) for tpg monitoring.

 1---
 2environment:
 3  job_name: monitoring-tpg-g5k-iotlab
 4  walltime: "00:59:00"
 5  g5k:
 6    cluster: paravance
 7    job_type: ["allow_classic_ssh"]
 8  iotlab:
 9    cluster: grenoble
10monitoring:
11  type: tpg
12  provider: g5k
13  cluster: paravance
14  network: shared
15  ipv: 6
16layers:
17- name: cloud
18  services:
19  - name: MyServer
20    environment: g5k
21    roles: [monitoring]
22    quantity: 1
23- name: edge
24  services:
25  - name: MyClientA
26    environment: iotlab
27    cluster: grenoble
28    archi: a8:at86rf231
29    roles: [monitoring]
30    quantity: 1
31  - name: MyClientB
32    environment: iotlab
33    cluster: grenoble
34    archi: rpi3:at86rf233
35    roles: [monitoring]
36    quantity: 1

Network Configuration

1networks:
2- src: cloud
3  dst: edge
4  delay: 180ms
5  rate: 1gbit
6  loss: 0.1%

Workflow Configuration

prepare installs stress on all Services.

launch runs stress on all Services.

 1# SERVER
 2- hosts: cloud.*
 3  prepare:
 4    - shell: apt install -y stress 
 5  launch:
 6    - debug:
 7        msg: "Running"
 8    - shell: stress --cpu 32 --timeout 30
 9      async: 30
10      poll: 0
11# CLIENT
12- hosts: edge.myclienta.*
13  prepare:
14    - shell: pip3 install stress
15  launch:
16    - debug:
17        msg: "Running"
18    - shell: stress -c 1
19      async: 30
20      poll: 0
21- hosts: edge.myclientb.*
22  prepare:
23    - shell: pip3 install stress
24  launch:
25    - debug:
26        msg: "Running"
27    - shell: stress -c 4
28      async: 30
29      poll: 0

Running & Verifying Experiment Results

Find below the commands to run this example.

$ e2clab layers-services ~/git/monitoring-tpg-g5k-iotlab/
                         ~/git/monitoring-tpg-g5k-iotlab/

$ e2clab workflow ~/git/monitoring-tpg-g5k-iotlab/ prepare

$ e2clab workflow ~/git/monitoring-tpg-g5k-iotlab/ launch

You can access the Grafana service to visualize monitoring data during experiment execution. You can access it as described in the following file ~/git/monitoring-tpg-g5k-iotlab/20231013-111406/layers_services-validate.yaml. See more details below:

'* * * * * * * Monitoring Service (started during workflow ''launch'' step)'
'Available at: http://localhost:3000'
'Access from your local machine: ssh -NL 3000:localhost:3000 paravance-40.rennes.grid5000.fr'
'username: admin / password: admin'
../_images/figure-cpu-g5k.png

Figure 1: CPU usage on G5K node ‘MyServer’.

../_images/figure-cpu-a8.png

Figure 2: CPU usage on FIT IoT a8 device ‘MyClientA’.

Wait at least one minute before finalizing the workflow and saving the monitoring data.

$ e2clab finalize ~/git/monitoring-tpg-g5k-iotlab/

The monitoring data will be saved at:

$ ls ~/git/monitoring-tpg-g5k-iotlab/20231013-111406/monitoring-data/
20231013T094209Z-7052d33e19ef5be9.tar.gz