Skip to content

0_setup.md

Latest commit

 

History

History
122 lines (81 loc) · 3.13 KB

0_setup.md

File metadata and controls

122 lines (81 loc) · 3.13 KB

Part 0: Setup and Prerequisites

In this step, we will verify that your system provides all necessary tools to run SeBS.

Verify Python Installation

Check Python version:

python --version

To use SeBS, you need Python 3.7 or newer.

Verify Docker Installation

docker --version

docker run hello-world

If you see a permission denied error:

sudo usermod -aG docker $USER

Then, try to run the hello-world command again in a new shell.

Install Required Tools

In the tutorial, we use jq to parse JSON outputs. Additionally, we recommend installing cURL dependencies to avoid issues when installing pycurl library.

First, check that jq is functional:

sudo apt-get update && sudo apt-get install -y jq

echo '{"name": "test"}' | jq '.name'

Expected output: "test"

On Ubuntu, these packages should be sufficient to install pycurl:

sudo apt-get install libssl-dev libcurl4-openssl-dev

Clone and Install SeBS

In this step, we will clone the SeBS repository and install it with support for all cloud and local platforms. SeBS will create a Python virtual environment to manage dependencies. After these steps, we will be ready to deploy benchmarks.

git clone --recursive https://github.com/spcl/serverless-benchmarks.git
cd serverless-benchmarks

Set Tutorial Directory Path:

Throughout this tutorial, we reference configuration files from the tutorial directory. Set this variable to make commands easier to copy:

export TUTORIAL_DIR=/path/to/sebs-tutorial

# Example: If you cloned the tutorial to ~/sebs-tutorial, use:
export TUTORIAL_DIR=~/sebs-tutorial

Install SeBS with local and OpenWhisk support

./install.py --aws --azure --gcp --local --openwhisk

Activate Python virtual environment

source python-venv/bin/activate

Verify SeBS installation

./sebs.py --help

Expected output:

Usage: sebs.py [OPTIONS] COMMAND [ARGS]...

Options:
  --help  Show this message and exit.

Commands:
  benchmark
  experiment
  local
  resources
  storage

Understand SeBS Architecture

SeBS has four main commands:

  1. benchmark: Deploy and invoke individual functions.
  2. experiment: Run systematic experiments (perf-cost, invocation overhead, container eviction, etc.)
  3. storage: Manage object storage (MinIO and ScyllaDB for local deployment or OpenWhisk)
  4. local: Manage local Docker deployments
  5. resources: Manage resources created in the cloud.

Here are a few essential concepts that will help you understand SeBS:

  • SeBS Workflow:
benchmark code → build code package / container → deploy to platform → invoke → measure
  • Caching: SeBS manages a cache in the cache directory to remember all resources created in the cloud, as well as cache all built code packages and containers.
  • Resource ID: each SeBS deployment uses a randomly generated resource ID to uniquely identify resources (functions, storage buckets, etc.) created in the cloud. When setting up a new cache directory in SeBS, you can provide a custom resource ID to reuse existing resources.