How Cloud Design Can Support Sustainability and Improve Efficiency

Sathish Krishnan
Published 12/22/2022
Share this on:

Supoorting sustainability with cloud designAs concerns about sustainability and climate change drive transformations in both the private and public sectors, more organizations and countries are adopting environmental goals to address their greenhouse gas emissions. International agreements like the Paris Climate Accords, which aim to limit the global mean temperature increase to 1.5 degrees Celsius, and the United Nations Sustainable Development Goals, which calls for countries to “take urgent action to combat climate change and its impacts,” have set the tone for emission reduction and sustainability. Additionally, many private-sector organizations, including the tech giants Amazon, Google, and Microsoft, have adopted their own sustainability goals to reach net-zero carbon emissions by 2040 or sooner.

Even smaller organizations can and should adopt sustainability goals to address their long-term environmental and economic impact. With technology playing an ever-increasing role in people’s daily lives and concerns about carbon emissions and climate change steadily growing, cloud computing is well-positioned as a less carbon-reliant alternative to traditional data centers. As cloud computing becomes the standard for industries across the board, several key design principles and best practices can help companies improve efficiency and support their sustainability goals.


Opportunities to improve sustainability with cloud computing

Data centers, whether traditional or cloud-based, require an enormous amount of energy to operate. One report from the International Energy Agency (IEA) puts global data center energy demand at nearly 200 terawatt-hours, or approximately one percent of the world’s electricity usage. It’s estimated that data centers may count for anywhere between two and three percent of global carbon emissions.

How data centers get that energy is important, and many are not operated using clean energy. For example, the state of Virginia is home to the world’s largest concentration of data centers in an area known as “Data Center Alley,” and an estimated 70 percent of global internet traffic passes through the area each day. But according to the Virginia Department of Environmental Quality, the state only generates about seven percent of electricity from renewable energy.

While companies and organizations can and should support public policies that increase the availability of clean energy, there are other opportunities to improve their own sustainability with cloud computing. Additionally, research from Virginia Tech and UC Berkeley shows that hyperscale cloud-based data centers are the most efficient type of data centers, generating only 0.15 tons of CO2-equivalent per computing workload compared to the 0.75 tons generated by internal data centers.



Want More Tech News? Subscribe to ComputingEdge Newsletter Today!



Considerations for cloud sustainability

Before initiating a cloud design project to support sustainability and improve efficiency, organizations should establish specific goals and ensure they are achievable, not only practically, but in terms of future growth and key performance indicators. While the ultimate goal is to reduce energy consumption, it’s important not to compromise performance.

During the design phase, it’s crucial to maximize utilization by right-sizing workloads, computing, storage, and networking. For example, imagine there are 10 nodes and each one runs at 30 percent utilization. Just two nodes running at 60 or 70 percent utilization could achieve the same goals.

The ability to optimize the networking layer is also key to achieving sustainability goals. When applications are available to customers around the world, the distance that packets must travel increases. Likewise, as packet size increases, more networking resources are required to transmit the data. Recently, a payment solutions provider for customers across the globe had end-users frustrated with the payment experience. The company did not want to add another region or data center to fulfill the customers’ needs; instead, it decided to utilize a content delivery network (CDN) and edge computing.

CDNs leverage edge locations around the world to cache content from origin servers. Delivering content from edge locations, which are closer to the end user, reduces the distance data travels. It’s important to monitor the CDN cache hit ratio, which determines whether the object is being served from the cache or must travel from the origin server. To reduce network traffic and enhance responsiveness, deploy application programming interface (API) payloads to edge-optimized API endpoints and cache the API response. CDNs can also be utilized to compress objects automatically, reducing the size of the data transmitted across the network. In the case of higher maximum transition units (MTU), use jumbo frames or packet fragmentation to enhance network performance.

Cloud Efficiency
Figure 1


Key design principles

Ultimately, designing a sustainable cloud computing system seeks to optimize efficiency and reduce unnecessary or unneeded storage, workloads, and operations. In doing so, energy consumption can be reduced, and the system simplified. Five design principles support this foundational goal.

  1. Reduce idle resources. Employ auto-scaling to automatically adjust computing resources based on central processing unit (CPU) utilization and, whenever possible, utilize general-purpose instances rather than high-performance or high-memory to save on costs. Moving to serverless, microservices, or container-based architectures can eliminate the need for physical servers and allow organizations to use only the computing needed. Using the management services of a large cloud provider passes the challenge of running the computer onto the cloud provider, permitting the customer to focus on ensuring business functionalities are maintained. Integrate smart tools such as instance schedulers and use scheduled auto-scaling to program shutdowns and terminate resources that run only during business hours or on weekdays. For flexible workloads, use spot instances.
  2. Optimize storage. Whenever possible, delete unused, excess data without compliance or audit requirements. Move infrequently used data to cold storage services where storage costs are low, but retrieval costs may be high. Optimize block and file storage by either downsizing or changing to an optimized volume to make transactions cheaper and run on more sustainable hardware. Data should be partitioned so that when queried, only a subset of the data is taken on. Transfer data in compressed formats to help reduce storage when data is moved across the network. Finally, instruct third parties to consolidate redundant data assets and remove unused and unnecessary storage.
  3. Use the most efficient instance types with the least impact. Many general-purpose workloads are not very busy and do not require a high level of sustained CPU performance. These low-to-moderate CPU utilization workloads lead to wasted CPU cycles, consuming more energy and costing companies more than necessary. Instead, start using burstable instances, especially when running a high-performance machine learning instance. Figure 2 illustrates CPU utilization for many common workloads that customers run in any public Cloud.

    Figure 2
  4. Optimize geographic placement of workloads. Knowing customer access patterns can help reduce the time it takes for network resources to reach the customer. When possible, use distributed data stores and caches closer to customer locations to avoid putting unnecessary load on the databases and eliminate the need for data to travel throughout the entire network.
  5. Keep software, hardware, and architecture updated. Upgrade operating systems, systems libraries, and applications to protect against security vulnerabilities and take advantage of regular sustainability improvements. Adopt new software offerings and architectural patterns, and move away from traditional processing toward more efficient architectures. Use asynchronous workloads and schedule large jobs during times of the day when the carbon cost of the power being used is the lowest.


Cloud computing for the future and improved finances

As long-term sustainability and carbon reduction continue to grow in importance, organizations need to identify ways to improve efficiency and decrease energy needs now. Implementing these strategies and best practices for cloud design helps companies establish a blueprint for sustainable development and position themselves for success in the face of increased public and governmental scrutiny.

The financial benefits of implementing sustainability goals can be significant, too. Optimizing cloud efficiency can lower energy costs and help organizations take advantage of newer and more efficient, and secure software and hardware offerings. With energy prices unlikely to decrease, organizations that begin to optimize their cloud workloads can establish and achieve their sustainability goals while reducing energy consumption and strengthening their bottom line.


About the Writer

Sathish Krishnan is a cloud infrastructure architect for Amazon Web Services with more than 12 years of experience in digital transformation, focusing on infrastructure, security, and machine learning. He can be reached at

Disclaimer: The author is completely responsible for the content of this article. The opinions expressed are their own and do not represent IEEE’s position nor that of the Computer Society nor its Leadership.