9 key considerations for reducing carbon emissions across your hybrid cloud estate
Recent reports from Gartner reveal that environmental IT sustainability strategy is a higher priority for CEOs than ever before. This shift in focus means organizations now need to capture and report on two main types of carbon emissions:
- Centralized corporate carbon emissions: Emissions resulting from centralized policies, such as a partnership with courier firms for shipping for the organization’s products, vehicle fleet usage, and employee travel policies.
- Department-specific carbon emissions: Emissions arising from day-to-day operational activities, which vary significantly across departments. For example, the logistics department would have a very different approach to measuring and reporting emissions compared to the information technology (IT) department.
There is bound to be overlap if software applications supporting corporate activities (e.g., expense reimbursement, HRMS) are hosted on-premises. In this article, we discuss capturing the information, so that these can be accurately reported under the desired headings.
For the Chief Information Officer (CIO) or the IT Director, the focus is likely to be on measuring and reporting on carbon emissions from their Hybrid Cloud Estate (a combination of on-premise and cloud landscapes). Important areas include:
- Energy consumption and efficiency
- Carbon footprint, including greenhouse gas emissions
- Data for Carbon Accounting of the technology landscape
Historically, decisions to migrate to the cloud or refresh on-premise data center hardware were primarily based on capital expenditure (cap-ex). However, as more information becomes available across the value chain of hardware, transportation, networking/data transfer and cloud, it is possible to overlay IT carbon emissions within this decision-making process. Carbon emissions are quickly becoming a crucial non-functional requirement (NFR) in these decisions.
Regulations such as Switzerland’s Ordinance on Mandatory Climate Disclosures, European Sustainability Reporting Standards (ESRS), UK Sustainability Disclosure Standards (SRS), and California’s Climate Accountability Package are making the baseline and remediation of carbon emissions a top priority.
Below, we discuss considerations for CIOs and IT Directors aimed at reducing carbon emissions within both on-premises and cloud infrastructures as related to their IT sustainability strategy, organized into three phases: Select, Operate and Dispose.
GreenOps phases
| Topic | On-Premise | Cloud |
| Select | ||
| Energy Efficiency – Resource
Reduce energy consumption where possible. |
Select hardware (servers and devices) with low power consumption ratings and low operating temperatures to minimize the overall energy requirements. | Select services which use low power compute to reduce the electrical energy consumed in the physical cloud region. |
| Energy Efficiency – Location
Minimize the carbon emissions from energy consumption where possible. |
In a multi-site data center, consider listing location by carbon emissions. Assess opportunities to grow the physical space available in the greenest location to be able to use this location to locate new servers. | Select regions for commissioning services which are either greener for electricity, or where the cloud service providers have power purchase agreements for green electricity. Note: this approach might create a multi-fold increase in emissions from data transmission. |
| Carbon Footprint – Embodied
Key non-functional requirement while assessing new hardware. |
Most manufacturers provide the carbon footprint as part of the device attributes. If the details are solely from manufacturer, it is important to overlay the likely emissions through the supply chain (packaging, shipping, delivery) to have an accurate figure. | Information is unavailable. However, that doesn’t mean that it doesn’t exist or that it can be excluded.
|
| Carbon Accounting – Amortization Period
Minimize the impact of embodied carbon.
|
Amortization of embodied carbon over the useful life. An accurate determination of the useful lifespan could prove advantageous in carbon accounting. Combined with usage emissions, this will provide a comprehensive figure for the period. | Information is unavailable and often the data available on emissions from public cloud infrastructure is solely emissions generated from usage. |
| Wider considerations | These aspects will need to be investigated to determine the overall carbon emissions from the on-premise IT assets:
|
Information is unavailable. However, these are important considerations for the cloud service providers as they grow the number of regions and services offered to their customers.
|
| Operate | ||
| Carbon Footprint – From Usage
Regularly look for opportunities to optimize the number of resources in the estate. |
Idle servers (and sometimes unused servers) across the application landscape need to be periodically reviewed to look at means of retirement or repurposing idle or unused infrastructure. | Idle servers (and sometimes unused servers) need to be periodically reviewed. These could be turned off to save costs as well as emissions. For cloud infrastructure, the recommendations from the cloud provider via the cloud consoles provide a good level of information. Also, commercially available FinOps solutions provide deeper insights into servers that need investigation. An existing Cloud FinOps team is potentially already looking at this from a cost reduction perspective. |
| Carbon Footprint – Usage
Regularly look for opportunities to leverage alternative architectures to meet peak load. |
Servers within on-premise applications are often sized for the peak load and hence are oversized (and emit higher carbon and operate at lower utilization). During the selection exercise, the possibility of looking into alternative architectures for the application to try to support peak load through other means could reduce the carbon footprint from usage. | Cloud compute reservations are a great commercial construct but not necessarily helpful for carbon footprint unless it is possible to operate these at higher utilization percentages. Opportunities to leverage horizontal scaling by applications to meet peak load requirements, as well as operate on smaller infrastructure at times of non-peak load, would reduce carbon footprint from usage. |
| Energy Proportionality – The more we utilize a computer, the more efficient it becomes at converting electricity to practical computing operations. | The best way towards energy efficiency is to run the workload on as few servers as possible, with the servers running at the highest utilization rate.
According to research by Dell, the power consumption for the PowerEdge R710 is summarized below:
As we see, the power consumption (and hence carbon emissions) is considerable even when the CPU isn’t loaded (i.e., at idle). Planning to maximize the CPU utilization of your organization’s servers and avoiding larger quantities of under-utilized servers would be a step towards overall carbon reductions. |
From the below estimate, the carbon emissions for one hour from operating at 100% utilization, is 18gCO2e, as compared to 8gCO2e when at 10% utilization.
Just like for on-premise infrastructure, it is important to ensure high utilization of compute infrastructure.
According to the Teads Engineering – EC2 Carbon Footprint Estimator: Instance: a1.2xlarge Region: US East (Ohio) Carbon emissions from one hour of computing
|
| Dispose | ||
| Reusing Resources | Flexera’s research shows that 35% of participants have processes to be able to donate some of these hardware assets at the end of its economic life. This is very useful in preventing more embodied carbon emitted to manufacture new devices. | This is limited to being able to leverage dedicated hosts for other purposes within the organization or repurposing reserved instances for use by another application if the initial reservation is not being used. |
| End of Life | Flexera’s research shows that 72% of participants have a defined sustainability policy or process for hardware assets. This is very useful for the circular economy and avoiding landfills. Decisions regarding recycling should also take into consideration the selection of replacement assets and the embodied carbon in those replacement assets. | Information is unavailable. However, that doesn’t mean that it doesn’t exist or that it can be excluded.
|
Conclusion
Baselining and remediation of carbon emissions from IT infrastructure—both cloud and on-premises—are now critical for organizations. With the availability of commercial solutions providing key data points for measuring IT carbon emissions, it’s imperative that information from manufacturers and cloud service providers becomes more widely accessible. Combining this data with standardized techniques and reporting standards will empower organizations to make data-driven decisions and effectively monitor and measure their IT assets’ carbon emissions.
Learn more about Flexera One FinOps.
References
