Introduction

GOXN - Green Observability eXperiment eNgine - is an extensible software framework to run green observability experiments and compare design decisions. GOXN follows the design principles of cloud benchmarking and strives towards portable and repeatable experiments. Experiments are defined as yaml-based configuration files, which allows them to be shared, versioned and repeated. GOXN automates every step of the experiment process in a straightforward manner, from SUE setup to data collection, processing and reporting. GOXN is a fork of OXN.

Installation

Prerequisites

• Kubectl + Kubernetes Cluster
• Python >= v3.10
• Jupyter

Running in kubernetes

Cluster Requirements

The cluster provides Persistent Volume Claims (PVCs) to store data over multiple pod restarts. For this, the cluster makes use of OpenEBS in the default given config of GOXN. Install OpenEBS with the following command:

kubectl apply -f https://openebs.github.io/charts/openebs-operator.yaml
kubectl patch storageclass openebs-hostpath -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

You can also use other implementations of PVCs. Just make sure to change the values in the helm configs accordingly.

Installgoxn via pip

Note:goxn requires Python >= 3.10

1. Install virtualenv

   pip install virtualenv

2. Create a virtualenv (named venv here)

   virtualenv venv

3. Source the venv

   source venv/bin/activate

4. Installgoxn

   pip install .

Note:goxn requires the pytables package, which in turn requires a set of dependencies.

Manual Execution and Deploxment

If you dont want to use the run_experiment_suite.sh script, you have to deploy the individual components using helm charts.

External Observability Stack

For the Prometheus and Grafana, we use the kube-prometheus-stack which deploys and configures Prometheus and Grafana in a ready to use state.

The following commands add the helm repository and install the kube-prometheus-stack in a specific namespace and apply custom configurations:

helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
helm repo update
helm install kube-prometheus prometheus-community/kube-prometheus-stack \
    --namespace oxn-external-monitoring \
    --create-namespace \
    --version 62.5.1 \
    -f values_kube_prometheus.yaml

Kepler also provides a Helm chart. Therefore, the provision is straightforward. We follow the instructions from the official documentation and execute the following commands. The Command deploys Kepler in the correct namespace and applies custom changes.

helm repo add kepler https://sustainable-computing-io.github.io/kepler-helm-chart
helm repo update

helm install kepler kepler/kepler \
    --namespace oxn-external-monitoring \
    --create-namespace \
    --set serviceMonitor.enabled=true \
    --set serviceMonitor.labels.release=kube-prometheus \
    -f values_kepler.yaml 

There is a preconfigured dashboard for Grafana. The dashboard is deployed from the GOXN repository using the command:

GF_POD=$(
    kubectl get pod \
        -n oxn-external-monitoring \
        -l app.kubernetes.io/name=grafana \
        -o jsonpath="{.items[0].metadata.name}"
)
kubectl cp kepler_dashboard.json oxn-external-monitoring/$GF_POD:/tmp/dashboards/kepler_dashboard.json

System Under Experiment Setup

Deployment of the SUE follows also the instructions from the official documentation Deploy the SUE in an own namespace and apply a custom configuration file:

helm repo add open-telemetry https://open-telemetry.github.io/opentelemetry-helm-charts
helm repo update
helm install astronomy-shop open-telemetry/opentelemetry-demo \
    --namespace system-under-evaluation \
    --create-namespace \
    --version 0.36.4 \
    -f values_opentelemetry_demo.yaml

Run an example observability experiment

1. Verify thatgoxn is correctly installed

oxn --help
usage: oxn [-h] [--times TIMES] [--report REPORT] [--accounting] [--randomize] [--extend EXTEND] [--loglevel [{debug,info,warning,error,critical}]] [--logfile LOG_FILE] [--timeout TIMEOUT] spec

Observability experiments engine

positional arguments:
  spec                  Path to angoxn experiment specification to execute.

options:
  -h, --help            show this help message and exit
  --times TIMES         Run the experiment n times. Default is 1
  --report REPORT       Create an experiment report at the specified location. If the file exists, it will be overwritten. If it does not exist, it will be created.
  --accounting          Capture resource usage forgoxn and the sue. Requires that the report option is set.Will increase the time it takes to run the experiment by about two seconds for each service in the sue.
  --randomize           Randomize the treatment execution order. Per default, treatments are executed in the order given in the experiment specification
  --extend EXTEND       Path to a treatment extension file. If specified, treatments in the file will be loaded intogoxn.
  --loglevel [{debug,info,warning,error,critical}]
                        Set the log level. Choose between debug, info, warning, error, critical. Default is info
  --logfile LOG_FILE    Write logs to a file. If the file does not exist, it will be created.
  --timeout TIMEOUT     Timeout after which we stop trying to build the SUE. Default is 1m

2. Run an experiment and write the experiment report to disk

oxn experiments/recommendation_pause_baseline.yml --report baseline_report.yml

Quick Start: Experiment Suite

sh run_experiment_suite.sh

This should automatically setup the needed helm charts and start the experiments. Make sure thatgoxn and its requirements are installed before.

Data Post-Processing

To create human readable insights, use the data_processing.ipynb located in evaluation/. This Jupyter Notebook creates easilier to understand CSV files. For even further processing, have a look at the steps described in the GOXN Replication Package.