Businesses are increasingly embracing data-intensive workloads, including high-performance computing, artificial intelligence (AI) and machine learning (ML). These technologies drive innovation on their hybrid, multicloud journeys while focusing on resilience, performance, security and compliance. Companies are also striving to balance this innovation with growing environmental, social and governance (ESG) regulations. For most organizations, IT operations and modernization form a part of their ESG objective, and according to a recent Foundry survey, about 60% of organizations seek service providers specializing in green technology areas.

As carbon emissions reporting becomes common worldwide, IBM is committed to assisting its clients in making informed decisions that can help address their energy demands and associated carbon impact while reducing costs. To aid in building more sustainable IT estates, IBM has partnered up with Amazon Web Services (AWS) to facilitate sustainable cloud modernization journeys.

As companies fast-track their IT modernization to accelerate digital transformation and gain business advantage, a significant opportunity emerges. This opportunity involves rearchitecting IT environments and application portfolios toward greener, more sustainable designs. Such an approach not only drives cost efficiencies but also contributes to broader corporate sustainability goals.

Understanding carbon emissions from digital technology

All business applications that IBM builds and runs, whether for external or internal customers, come with a carbon cost, which is primarily due to electricity consumption. Regardless of the technology that IBM used to develop these applications or services, operating them requires hardware that consumes power.
The carbon dioxide (CO2) emissions produced by grid electricity vary based on the generation methods. Fossil fuels such as coal and gas emit significant amounts of carbon, whereas renewable sources such as wind or solar emit negligible amounts. Thus, each kilowatt (kW) of electricity consumed directly contributes to a specific amount of CO2 equivalent (CO2e) released into the atmosphere.

Therefore, reducing electricity consumption directly leads to lower carbon emissions.

Carbon footprint in practice

Compute, storage and networking are the essential tech resources that consume energy in the process of building applications and services. Their activity requires active cooling and management of the data center spaces that they operate in. As custodians of sustainable IT practices, we must consider how we can reduce the consumption of resources through our daily activities.

Figure 1: Data centers require electricity to power core IT resources such as compute, storage and networking

Data centers draw power from the grid that supplies their operational region. This power runs various IT equipment such as servers, network switches and storage, which in turn support applications and services for customers. This power also operates ancillary systems such as heating, ventilation and air conditioning or cooling, which are essential for maintaining an environment that keeps the hardware within operational limits.

A way forward to decarbonization

Modernizing applications is becoming pivotal for driving innovation and transforming businesses. IBM Consulting® applies the AWS Well-Architected framework to create a Custom Lens for Sustainability to perform workload assessments for applications both on premises and on AWS Cloud. To read about other key scenarios and entry points of IBM Consulting® Custom Lens for Sustainability, check out the blog post: Sustainable App Modernization Using AWS Cloud.

In this blog post, we delve into an in-depth analysis to assess, implement recommendations on, and analyze the carbon emission effects of a monolithic application running on AWS through a sustainability lens.

Green IT Analyzer: A comprehensive IT decarbonization platform

The Green IT Analyzer platform enables clients to transform their traditional IT into more energy-efficient, sustainable green IT. Serving as a one-stop shop, it measures, reports, creates baselines and provides a unified dashboard view of the carbon footprint across the hybrid cloud environment—including private data centers, public cloud and user devices. The platform can measure the carbon footprint of the IT estate at both a granular and virtual machine (VM) level. It helps identify energy or carbon hotspots to develop an optimization roadmap. The carbon assessment technique that it uses aligns with greenhouse gas (GHG) principles for the information and communication technology sector.

Figure 2: Green IT Analyzer platform, an IBM asset available on AWS Cloud

Location-based methodology

Understanding the carbon emissions from IT workloads requires familiarity with several key concepts and metrics. Here’s a high-level overview:

Figure 3: Methodology to distribute energy from physical to logical layer
  • Carbon footprint (CFP): The concept of carbon footprint is central to our analysis. CFP represents the total amount of CO2 and equivalent GHG emissions associated with powering a data center, starting from a baseline measurement of CFP greater than or equal to zero. It’s a crucial metric for gauging the environmental impact of data center operations.
  • Power usage effectiveness (PUE): Another critical metric is power usage effectiveness. PUE measures the energy efficiency of a data center, calculated by dividing total facility energy by the energy consumed by IT equipment. This division yields a ratio that indicates efficiency: a PUE close to 1 (one) signifies high efficiency, while higher values suggest greater energy waste.
    Formula: PUE = (total facility energy)/(energy consumed by IT equipment)
  • Carbon intensity (CI): Lastly, we consider carbon intensity. CI measures the carbon emissions in grams per kilowatt-hour (g/kWh) of grid power generation that powers the data center. This metric varies based on the energy source. Coal-powered grids can have a CI that is greater than 1,000 g/kWh while grids powered by renewable sources such as wind and solar should have a CI closer to zero. (Solar panels have some embodied CFP but have much less compared to fossil fuels.)
Figure 4: Distribution of energy consumed from electricity grid to physical equipment and then virtualized layer

Let’s consider a major client challenge. Every organization is committed to achieving net-zero emissions, and IT plays a crucial role in achieving the sustainability agenda. This can involve reducing the carbon footprint of the IT estate itself—especially relevant for financial customers with high IT-driven emissions—or creating a sustainable platform that runs on green IT.

Older monolithic applications, typically running on VM-based platforms in either on-prem data centers or public clouds, are a key focus area. A crucial question arises: how can we reduce IT resource consumption from these older monolithic applications, which generally hold 20–30% of the entire IT portfolio? It is more energy-efficient to move from VM-based monolithic applications toward a more energy-efficient, microservice-based architecture running on a container platform. However, it’s essential to evaluate each case individually, as a one-size-fits-all approach is not always effective.

This criteria can be used to select application transformation candidates:

  • Applications with more than 70%–80% CPU utilization
  • Applications experiencing seasonal spikes in transactions, such as around Christmas Eve, Diwali and other public holidays
  • Applications with daily spikes in transactions at specific times, such as airline onboarding in the early morning or night
  • Some business components within monolithic applications that exhibit usage spikes

As-is state analysis of monolithic apps

Consider the example of a simple e-Store application running on AWS in an Elastic Compute Cloud (EC2) VM. This application, an e-CART, experiences seasonal workloads and has been rehosted (lift-and-shift) from on premises to an AWS EC2 instance. Monolithic applications like this package all business functions into a single deployable unit.

Figure 5: Monolithic e-CART application architecture 

The following table describes the key characteristics of e-Store legacy applications.

Application characteristicsName or identifiere-Store Application
 Runtime and versionsJDK 8
 OS and environmentsNo. of production instances: 1; OS: Ubuntu; Env: Dev, Test, UAT, Prod, DR
 TechnologiesJSPs, Servlets, Spring Framework, Log4j; no caching and session management
Databases characteristicsDatabaseDatabase: 1; growth rate: 10% year-over-year
Operational characteristicsServer capacityt2.large Database: 32GB RAM with 75% utilization; vCPUs: 2; storage: 200GB
 Availability zoneUs-east-1d
 NFRsAmount of total users: 10,000; Amount of concurrent users: 500; Types of users: Internal; TPS: 100; Peak usage period: First week of the month; Uptime: 99%; Performance: Page should be loaded within 2 seconds; Security classification: CIA-M/H/H; Regulatory requirements: None; Monitoring: Manual health checks; DevOps: Git and Jenkins
Scroll to view full table

The carbon emissions of a workload are directly linked to the consumption of resources such as computing, storage and network, with computing often being the most significant contributor. This varies based on workload characteristics; for instance, in the media or streaming industry, data transmission over the network and storing large unstructured data sets consume considerable energy.

The graph shows the utilization pattern of the CPU when minimal user activity is happening on the monolithic application running in a single EC2 instance.

Figure 6: CPU utilization of VMs with minimum transactions over a period of time

We used the Green IT Analyzer platform to conduct a carbon accounting of the as-is state of the monolithic application, comparing it to the target state of the same application when rearchitected into a microservice architecture running on the Amazon Elastic Kubernetes Services (EKS) platform.

Step 1: Comprehensive carbon footprint analysis of monolithic applications

First, we focus on examining the current carbon footprint of a monolithic workload under various operating conditions. This provides us with a baseline for identifying areas for improvement.

Let’s calculate the estimated carbon footprint for our monolithic workload when we have minimal user transactions and 45% of CPU utilization:

  • PUE of US east 1d AZ: 1.2
  • CI: 415.755 grams of CO2/kWh

A. Estimated carbon calculation when there is no user activity:

  • Energy consumed: 9.76 g/W @ 45% utilization
  • Hours of running the same workload: 300 hours
  • Estimated carbon emissions for 300 hours = PUE × CI × energy consumed by workload
  • = [(1.2 × 415.755 × 9.76) × 300] ÷ 1,000 = 1,460.79 grams of CO2e

B. Estimated carbon emission with concurrent 500 users:

In a scenario where peak-level transactions were created as per non-functional requirements (NFR) to test the system’s ability to support daily peaks, CPU utilization surged to 80% during concurrent user activity. This situation triggered an auto-scaling rule set to activate at 80% CPU utilization. The rule provisions extra VMs to help ensure that the load on each VM remains below 60%. The load balancer then efficiently distributes the load among both the existing and new VMs.

Due to the auto-scaling of the new EC2 instances, an additional t2.large VM became available, which led to a drop in the average utilization to 40%.

  • Estimated carbon emissions for this scenario, with both identical VMs running for 300 hours = PUE × CI × energy consumed by workload
  • = {[(1.2 × 415.755 × 9.76) × 300] × 2} ÷ 1,000 = 2,921.59 grams of CO2e

Step 2: Implementing sustainability recommendations

This step explores a range of sustainability recommendations and their practical implementation for the monolithic application. We use the Custom Lens assessment for Sustainability to guide these recommendations.

First, we consider decomposing monolithic applications into action-based reactive microservices. This approach is tailored to the application’s seasonal behavior and varying usage patterns, which is particularly useful during peak periods such as festive seasons when traffic surges and a focus on browsing artifacts over backend transactions is observed.

Second, the plan involves reducing energy consumption by scheduling batch processing during idle periods, especially when the data center grid operates on green energy. This approach aims to conserve power by minimizing the duration of long-running transactions.

Finally, the strategy emphasizes the importance of choosing a flexible platform, such as AWS EKS or Red Hat® OpenShift® on AWS (ROSA), that is capable of dynamically scaling resources based on network traffic. Such a platform choice helps ensure optimized resource allocation and is beneficial for hosting the action-based reactive microservices.

In summary, the proposed strategies include microservice decomposition aligned with usage patterns, energy-conscious transaction scheduling, and a flexible platform choice to enhance application efficiency and resource utilization.

The application refactored into microservices is shown in the image:

Figure 7: Monolithic application decomposed into 4 microservices

Now let’s calculate the carbon emission after transforming the monolithic application to microservices-based architecture following sustainable design principles while refactoring the application under the umbrella of sustainable modernization.

A. Estimated carbon accounting with no or few loads:

  • Worker node: 2 × t2.medium
  • Utilization: 10% (when there is no load on the application)
  • Energy consumed: 6 g/W at 5% utilization
  • PUE (1.2) and CI (415.755 grams of CO2/kWh) remain the same because we are continuing to use the same availability zone.
  • Hours: 300
  • Estimated carbon emissions for 300 hours = PUE × CI × energy consumed by workload
  • = [(1.2 × 415.755 × 6) × 300] ÷ 1,000 = 1,796 grams of CO2e

Observations: When there is no load on the system, an application running on a VM is more carbon efficient than microservices running on an EKS cluster.

B. Estimated carbon accounting during peak load:

Similar to the load testing of monolithic applications, we onboarded 500 users and triggered concurrent transactions to meet the NFR requirements in the microservices that we built.

  • Worker node: 2 × t2.medium
  • Increased utilization due to load: 10% to 20%
  • Energy consumed: 7.4 g/W at 20% utilization
  • PUE and CI remain the same.
  • Hours: 300
  • Estimated carbon emissions for 300 hours = PUE × CI × energy consumed by workload
  • = [(1.2 × 415.755 × 7.4) × 300] ÷ 1,000 = 2,215.14 grams of CO2e

Here, autoscaling of pods occurred for UI services, but cart services did not require more resources to scale up. In monolithic applications, scaling up the entire platform is necessary regardless of which business functions or services require more resources, leading to increased utilization of 20%.

Observations: Let’s compare both scenarios.

  1. When the system is idle or has a steady load profile across the clock: When there is almost no load, monolithic applications consume fewer resources and emit nearly 18% less carbon than microservices-based applications hosted in the EKS cluster.
  2. When the system is on full load or varying load: When the system is on full load, there is a 24% reduction in CO2 emissions on the Kubernetes platform compared to a VM-based workload. This is due to the use of fewer cores and lower utilization. We can move more workloads in the same cluster and free up more cores from other applications to get more significant benefits.
Figure 8: Carbon emissions pattern of different architectural styles

This scenario is an example of how IBM® Custom Lens assessment for Sustainability on AWS workload helps to design your sustainable modernization path and reduce the total carbon footprint of your IT estate.

Action guide

For organizations that value sustainability, responsible computing and green IT are not just vital; they are entirely feasible. IT leaders can achieve these goals by pursuing environmentally friendly activities that encompass IT strategy, operations and platforms.

  • Greening your IT platforms: Use refactoring to migrate applications to the public cloud. Migrating workloads to the public cloud without optimizing them for this environment can increase operating costs and reduce sustainability. Instead, enhance workloads to be more cloud-native by refactoring applications based on factors such as their lifecycle, update and deployment frequency, and business criticality.
  • Optimizing idle VM capacity and other unused cloud resources: Enable infrastructure-level observability to identify idle VMs across your IT estate. Implement rules-based automation to take corrective actions, such as deleting idle VMs and associated resources that no longer serve business functions. Additionally, optimize VM sizing based on network traffic through auto-scaling.
  • Creating resources when needed: Although cloud resources are elastic, you gain limited efficiency benefits if you deploy workloads to fixed resources that run continuously, regardless of usage. Identify opportunities to provision and delete resources as needed, such as using VM scheduling or elastic features within cloud services.
  • Containerizing workloads: By using a container platform instead of a traditional VM environment, you can reduce annual infrastructure costs by up to 75%. Container platforms allow for efficient scheduling of containers across a cluster of VMs based on their resourcing requirements.
  • Modernizing your monolithic applications to microservices-based architecture: Select reactive microservices based on your needs: reactive microservices for event-based invocation to optimize resource utilization, event-driven microservices for asynchronous invocation, or serverless microservices for need-based execution of a single function.

The IBM Consulting Green IT Transformation framework, Custom Lens for Sustainability, and the Green IT Analyzer platform collectively help clients on their decarbonization journey. Both frameworks help assess workloads, identify optimization levers that can lower energy consumption, and create an application modernization roadmap that enables you to achieve your sustainability goals.

Learn more about IBM Consulting services for AWS Cloud.
Was this article helpful?

More from Cloud

The history of the central processing unit (CPU)

10 min read - The central processing unit (CPU) is the computer’s brain. It handles the assignment and processing of tasks, in addition to functions that make a computer run. There’s no way to overstate the importance of the CPU to computing. Virtually all computer systems contain, at the least, some type of basic CPU. Regardless of whether they’re used in personal computers (PCs), laptops, tablets, smartphones or even in supercomputers whose output is so strong it must be measured in floating-point operations per…

A clear path to value: Overcome challenges on your FinOps journey 

3 min read - In recent years, cloud adoption services have accelerated, with companies increasingly moving from traditional on-premises hosting to public cloud solutions. However, the rise of hybrid and multi-cloud patterns has led to challenges in optimizing value and controlling cloud expenditure, resulting in a shift from capital to operational expenses.   According to a Gartner report, cloud operational expenses are expected to surpass traditional IT spending, reflecting the ongoing transformation in expenditure patterns by 2025. FinOps is an evolving cloud financial management discipline…

IBM Power8 end of service: What are my options?

3 min read - IBM Power8® generation of IBM Power Systems was introduced ten years ago and it is now time to retire that generation. The end-of-service (EoS) support for the entire IBM Power8 server line is scheduled for this year, commencing in March 2024 and concluding in October 2024. EoS dates vary by model: 31 March 2024: maintenance expires for Power Systems S812LC, S822, S822L, 822LC, 824 and 824L. 31 May 2024: maintenance expires for Power Systems S812L, S814 and 822LC. 31 October…

IBM Newsletters

Get our newsletters and topic updates that deliver the latest thought leadership and insights on emerging trends.
Subscribe now More newsletters