Introduction

This short presentation aims at explaining the why behind the creation of Boavizta cloud-scanner.

This project started out of a Boavizta hackathon, initially as a way to demonstrate that we could make good use of Boavizta Open data and Open methodology to help assess environmental impacts of public cloud usage.

Cloud scanner evolved as a open source project, part of the Boavizta ecosystem. It offers a way to address some of the challenges we identified with de default environmental reporting of most public cloud providers.

We will first explain why we are not satisfied with the status-quo, then explain the Boavizta/cloud-scanner approach.

We will also highlight the limits of our approach and the roadmap of future cloud-scanner.

This presentation addresses the why. But if your are more interested by the how feel free to jump directly into cloud-scanner documentation.

What we need

To assess environmental impacts of public cloud resources, we need:

A global view of the environmental cost

  • Aggregated impact data
  • Use phase + manufacture phase
  • Multiple criteria (GWP, ADP...)

Real time monitoring

  • greenops approach
  • notification / alerting
  • trigger other processes

Challenge alternative architectures

  • Need to simulate variant of cloud architectures

Challenges with cloud providers

We face several challenges with the default environmental reporting of public cloud providers.

A lack of transparency

  • Questionnable methodology (not transparent / not documented)
  • Scope (1 and 2 only ?)
  • Decried Market-based approach 💣

A granularity problem

  • in time (3 months delay)
  • in service (hard to break down by app or micro service)

No easy integration

  • difficult to reuse the data

  • cf AWS dashboard.

Our approach

Transparent

  • Open and public methodology (peer reviewed)
  • Open data set of impacts
  • Boavizta cloud scanner is open source

Real time

  • 5 minutes sampling period

Fine grain attribution (tags)

Integration

  • generating metrics (prometheus)
  • exporting capabilities
    • files
    • metrics

Simulation

  • capacity to extract or inject a cloud inventory

Limits

Does not cover all aspects

  • instances and storage only
  • no serverless (lambda)
  • not storage object (S3)
  • not the entire scope (control plan + cloud infra not accounted)

=> underestimate !

Attribution for Kubernetes

  • fine grain attribution in case of K8S
  • CF work of SBS.

Roadmap

More services

A first objective is to increase the scope of the services.

  • Object storage / S3
  • FAAS/Serverless (like AWS lambda or Azure functions)
  • Additional managed services (like DB as a service)
  • impacts of GPUs, which is key to assess AI/ML workload(!)

For each new service, this will likely be a 2 step process:

  1. Define a common model (identify key drivers of the impacts) and implement this model in Boavizta API.
  2. Automate collection of related data in cloud scanner.

Support additional providers

  • Azure (work in progress)
  • Scaleway (work in progress)
  • OVH
  • ...

Providing recommendations

Estimating environmental impacts only makes sense if we can use the results to regain the ability to avoid unnecessary impacts.

In the future cloud-scanner or another project may provide insights on what is wasted and were to apply efforts of reduction.

Frequently asked questions

Isn't the reporting offered by cloud provider sufficient ?

For several reasons (scope, transparency, granularity), we consider that today's reporting of major public cloud provider does not fully answers to our needs. See Challenges and What we need.

A comparison of what is reported by default is addressed with more detail in this Boavizta article: Understanding the results of cloud providers' carbon calculators | Boavizta.

You may also refer to the academic paper that describes the Boavizta cloud methodology: BoaviztAPI: a bottom-up model to assess the environmental impacts of cloud services.

References

Boavizta

Repositories / tools

Methodology

Other projects that we find interesting

Some of these projects directly contributed to our work or are inspirations for the future development of cloud-scanner. Thank you !

Contributing

Evaluating the environmental impacts of public cloud resources is a work in progress. We are convinced that it involves a multidisciplinary approach and open discutions.

It is also only a first step towards a better use of the resources.

Reach out

Try cloud scanner, open issues on Github or join discussions on Github or Boavizta Mattermost channel if you have questions of would like to improve it.

Thank you !

Any question ?