Sustainable digital infrastructure alliance (SDIA)

Scope

The Sustainable Digital Infrastructure Alliance (SDIA) is a non-profit independent alliance of stakeholders that focuses on promoting and developing sustainable practices in the digital infrastructure industry. Its primary aim is to create a sustainable, scalable, and efficient digital infrastructure (data centres, cloud computing, and networks) ecosystem that supports ongoing digital transformation while minimising environmental impact. SDIA fosters collaboration (more than 65 members) among industry stakeholders, including technology companies, data centre operators, policymakers, and research institutions, to connect, benchmark, and deliver sustainability within their Digital Infrastructure value chain segment. SDIA raises awareness about the importance of sustainability in digital infrastructure and provides educational resources to promote best practices while ensuring they remain open and accessible to everyone.

Summary

Building on the Life Cycle Assessment (LCA) methodology, SDIA aims to extract insights into the environmental impact of digital products and services. By establishing a clear taxonomy of the value chain, adopting an innovative approach to digital resource primitives, and converting these digital resources into environmental impact metrics, we move closer to accurately assessing the environmental footprint of digital products and services.

Taxonomy[1]. Software applications have a supply chain known as digital infrastructure. This infrastructure generates digital resources necessary to power the application and provides connectivity, allowing users to access the application. With browsers, users can connect to an application using devices like smartphones or computers. While the application is not a physical product, it relies on physical-digital resources to operate. These resources are supplied by digital infrastructure, which includes IT hardware, networking, and data centre facilities. The primary input in this supply chain is electricity. Since digital infrastructure generates heat, we refer to the input as energy encompassing electricity and heat. The main output of this supply chain is digital resources.

Responsibilities. The responsibility for the environment sustainability of a digital product is shared among multiple stakeholders across the value chain as show in the figure below.

Digital Resource Primitive. Digital Resource Primitives are defined as the low-level resources required for digital products and services to operate. They can be seen as the fuel that powers software applications. Given that multiple digital products may be enabled by the same digital infrastructure, after identifying digital resource primitives utilised by the digital product, the weighting that identifies the utilization of the resource should also be assigned, as presented in the example of the figure below.

The indicators utilised for LCA environmental impact of different resource primitives available are:

  • Climate change – total, fossil, biogenic and land use → kg CO2-eq
  • Depletion of abiotic resources – minerals and metals → kg Sb-eq
  • Water use → m3 world eq. deprived
  • Land use → Dimensionless
  • Primary renewable energy (energy) → MJ
  • Primary non-renewable energy (total) → MJ

Additional indicators and parameters are available [1, 2, 3]. The complete list of environmental impact indicators, resource usage indicators, waste indicators, and output flow indicators is available in the SDIA Knowledge Hub [4]. Also, an example of an LCA methodology application for a webpage is available in [1].

Energy use formulas. In environments where physical energy measures are not available, e.g., through RAPL or IPMI, SDIA suggests a set of formulas for a best-effort estimate [5].

Tools and APIs [6]. SDIA categorises the tools and APIs into two dimensions: the ones that help assess the embodied impact of the equipment and infrastructure (static) and those that help determine the operational impact stemming mainly from energy and water usage (dynamic). A set of APIs for retrieving the embodied impacts of a physical server and an AWS instance is available for the first category. For the operational impact, there are available tools for measuring the impact on physical and virtual machines using different methods based on SDIA formulas and machine learning models, as well as models for estimating the power usage of a CPU.

Environmental Labels for Digital Resources [7]. IT Sustainability is becoming a strategic priority for large organisations globally. However, many face challenges obtaining the necessary information to generate reports and identify key focus areas. Procurement can significantly contribute by demanding transparent environmental impact data from IT suppliers, including cloud infrastructure, hosting, software, and services, and by setting clear criteria for digital resources. The SDIA’s Type I Sustainability Certification for digital resources aids procurement teams in acquiring accurate information and sustainably-produced resources from suppliers. This certification service is under development and aims to create labels based on sustainable cloud infrastructure (CSC) criteria.

The SDIA will serve as an independent platform for academia, technical experts, and suppliers to collaborate and advance a set of ambitious criteria that should qualify a cloud or IT infrastructure provider to receive a label for a digital resource product. After ensuring that the proposed criteria meet the SDIA’s values and roadmap, TCO Development will be the qualified certifier to develop the implementation with suppliers.

Datacenter metrics [8]. The SDIA knowledge Hub includes a set of metrics and formulas for estimating the energy consumption, efficiency, and carbon emissions of data centre infrastructure as a whole and for individual pieces of IT hardware based on ISO and EN standards.

  • The total energy consumption (EDC) will be measured as the energy consumption of the supporting equipment (cooling, lighting, etc.) and the IT hardware.

N is the total number of energy consuming objects within the data center facility. Ei is the amount of energy consumed by an object i.

ERen represents the total amount of energy coming from renewables in kWh. EDC represents the total amount of energy consumed by the data centre in kWh.

  • On-site Energy Fraction (OEF). The OEF sets the share of renewables for energy procurement concerning on-site energy production. While the carbon intensity of the local energy grid fluctuates and cannot be influenced by the data centre, data centres with on-site energy production have a source of 100% renewable energy. The OEF calculates how much of the required power is facilitated by the on-site renewables. Ideally, the renewables cover 100% of the load power the data centre consumes, equaling an OEF value of 1.

R(t) represents the renewable power produced at time t; L(t) is the load power consumed at time t; dt represents the time period of the calculation.

  • Power Usage Effectiveness (PUE) is the most well-known data centre efficiency metric currently and is defined in ISO/IEC 30134-2 / EN 50600-4-2. It calculates the facility’s total energy consumption in relation to the energy used for IT.
  • IT Usage Effectiveness (ITUE). Although PUE is the most popular metric in terms of energy efficiency, it does not represent the efficiency of the actual IT. To take this into account, Intel has developed the ITUE metric, adding one layer of detail to the energy consumption of IT Equipment. The ITUE is a PUE-type metric for IT equipment rather than for the data centre.
  • Energy Reuse Factor (ERF). The ERF, defined in ISO/IEC 30134-6 / EN 50600-4-6, determines the share of total energy consumption that is reused.
  • Energy Reuse Effectiveness (ERE). Energy reuse effectiveness—which is not part of the standard—can be calculated from PUE multiplied by the Energy Reuse Factor.
  • Carbon intensity of total energy consumption. Once the energy data has been collected, we can derive the carbon emissions from that number using local energy grids’ carbon intensity information. gain, to quantify the total emissions unrelated to efficiency, we measure the total carbon emissions from the data centre. Therefore, we will multiply the total energy consumption by (1- OEF, On-site energy fraction) to get the total energy consumption that comes from the energy grid. Additionally, we reduce this number by the amount of energy being reused within the facility.

With this calculation, we get the net energy consumption of a data centre fed by the grid. We then multiply this number by the CO2-equivalent of the local energy grid, which is generally available within 24 hours after the actual energy consumption. We then add the CO2 Emissions of the backup generator in case it is running on fossil fuels.

CO2DC is the data centre’s carbon emissions in kg. EDC is its total energy consumption in kWh. OEF is the On-site energy fraction. Ereuse is the energy being reused in kWh. CO2eq, Grid is the carbon intensity of the local energy grid in kg/kWh.

CO2UPS is the carbon emissions of the backup generator.

  • Carbon emission factor (CEF). From the above data, we can also calculate the carbon emission factor. For this metric, the total carbon emissions are divided by the total energy consumption. Effectively, this metric will return the grid’s carbon intensity, so it needs to be discussed whether we will implement it.

CO2DC is the total carbon emissions by the data centre in kg. ET is the total energy consumed by the data centre in kWh.

  • Carbon Usage Effectiveness (CUE). Carbon Usage Effectiveness (CUE) is described in ISO/IEC 30134-8 / EN 50600-4-8.

The draft proposes 3 categories to assess and report carbon emissions:

    • Category 1 (basic): accounting of external and internal DC electricity reported in CO2 emissions.
    • Category 2 (intermediate): accounting for external and internal DC electricity, all additional DC energy supply, and all additional DC emission sources, which are reported in CO2 equivalents.
    • Category 3 (advanced): reserved for future use.
  • Amount of useful energy consumed.
  • Server energy effectiveness metric. ISO/IEC 21836 defines a metric to measure the energy effectiveness of servers. It is based on SPEC SERT™v2 and defines the conditions under which the effectiveness of servers can be rated. A similar approach is undertaken by the European standard ETSI EN 303 470. The metric for the active state is:
  • Where WCPU, WMemory and WStorage are the weightings applied to the CPU, Memory and Storage worklets respectively. While the weightings are not defined in ISO/IEC 21836, ETSI EN 303 470 defines the weightings as follows:
    • WCPU = 0.65
    • WMemory = 0.30
    • WStorage = 0.05
  • With this approach, every server under inspection is characterised by a number and can be compared to other servers.

References

[1]    SDIA Report, “Creating a digital environmental footprint: a Life Cycle Assessment approach”, 2022. 

[2]    SDIA Report, “Commentary on the Energy Efficiency Directive”, 2022. 

[3]    SDIA Document, “List of impact indicators for software”.

[4]    SDIA Knowledge Hub, “Digital Environmental Footprint”.

[21]    SDIA Knowledge Hub,  “Digital Environmental Footprint – energy use formulas”.

[22]    SDIA Knowledge Hub, “Overview Tools & APIs

[23]    SDIA Knowledge Hub, “Environmental Label for digital resources” 

[24]    SDIA Knowledge Hub, “Datacenter metrics

Relevance for EXIGENCE

The work of SDIA is relevant to activities in relation to the overall approach for LCA and introduction of the concept of digital resources and labels, as well as for energy metrics and energy measurements for metrics definitions, formulas and tools.

Index