Cloud Sustainability Reporting

Introduction

Cloud sustainability metrics are available within Cloudability reports and dashboards and Apptio BI. Cloudability users can quickly and easily view the cloud sustainability metrics to get insights into their public cloud carbon emissions by using the below metrics:

• Estimated Carbon Emissions (MTCO2e) – This metric captures the estimated carbon emissions in metric tons of carbon dioxide equivalents.

• Power Consumed (kWh) – Power consumed in kilo watt hours. Cloudability calculates the cloud carbon footprint for all major CSPs (AWS, Azure, GCP & OCI) and offers a uniform methodology for comparing cloud carbon emissions across different vendors.

These metrics are available at a resource level for the below supported services.

Where to find the metrics

Cloudability > New Reports or Add Widget > Sustainability category

Apptio BI > New Report > Under Datasource and Measure select Cloudability Cost & usage > Sustainability category

Before diving into the methodology of these calculations, let’s understand the various emission types. Carbon footprints are generally originated from two primary sources: operational and embodied emissions.

• Operational emissions : Operational emissions are the result of day-to-day operation of cloud infrastructure. These emissions are generated primarily from the electricity consumed by cloud data centers to power and cool the servers, networking equipment, and storage devices that host cloud services.

• Embodied emissions : These are greenhouse gases released over the entire lifecycle of a product or service. This includes emissions generated during its manufacture, transportation, and eventual disposal.

Emissions have been further categorized into:

• Direct emissions : Emissions released directly from sources owned or controlled by the organization.

• Indirect emissions : Emissions that occur from the activities of the organization but are generated from sources not owned or directly controlled by the organization.

Carbon emissions are classified into various scopes:

• Scope 1 ( direct emissions ): These are the emissions generated from sources that are owned or controlled by the company. Here are the examples:

  • Fuel combustion in company-owned power backups
  • On-site manufacturing processes
  • Company-owned boilers and furnaces

• Scope 2 ( indirect emissions ): These are the emissions generated from the purchase of energy sources such as electricity and consumed by the company. Here are the examples:

  • Purchased electricity for office buildings
  • Purchased steam/ gas for industrial processes

• Scope 3 ( indirect emissions ): These are the emissions that are a consequence of the company's activities but occur within the supply chain. Here are the examples:

  • Operational emissions occurred by the use of the services provided by CSPs, such as virtual machines and storage among others.

As a customer of cloud services, we use the services offered by CSPs without any control over the energy generated to run the cloud infrastructure. Therefore, the operational emissions resulting from customer's cloud usage fall under scope 3 emissions. Currently, Cloudability considers only Scope 3 operational emissions and publishes the same via the sustainability metrics.

Methodology

The methodology of Cloudability ’s cloud sustainability metrcis is based on Cloud Carbon Footprint (CCF), with added enhancements like using actual utilization details of a resource and calculating power consumption via machine learning techniques.

In the methodology, we are using a bottom-up approach where the emissions of individual resources are calculated separately. These calculated emissions can be aggregated to collectively represent the an organisation's emission due to its cloud infrastructure usage.

Formula used to calculate the CO2e emission of cloud:

Estimated Carbon emissions = Power Consumption per hour x Power Usage Effectiveness (PUE) x Grid Emissions x Usage Hours

Where:

• Power consumption per hour: Use machine learning to predict the instance's hourly power consumption, using training data that includes machine specifications (vCPUs, memory in GB etc) and utilization metrics.

• Power Usage Effectiveness (PUE): Metric to measure the energy efficiency of a data center by comparing the total energy consumption of the facility to the energy used by the IT equipment.

• Grid emissions: Amount of carbon emissions per unit of electricity generated at a yearly level.

• Usage Hours: Total operating hours of the machine.

Data sources used for these:

Benefits of this Methodology

• It provides a uniform methodology that allows for comparison of cloud carbon emission across different CSPs (AWS, Azure, GCP and OCI).
• Power consumed is calculated by a model which provides power predictions for each resource based on its actual utilization.
• Machine learning model and calculation are validated by IBM Research Labs.

Prerequisites

To access the Cloud Sustainability metrics, customers should meet the following prerequisites:

• For accessing cloud carbon metrics, customers must have advanced credentials enabled for calculations at actual utilization level, else we'll default to 50% utilization for calculations.
• For GCP, Detailed billing must be enabled.

Exclusions

• GCP custom machines sustainability metrics are not available.
• Embodied emissions are not part of the calculations currently.