by Phil Goldstein (a cloud migration report translated from the AWS Blog – original here)

Moving compute, storage, and networking to the cloud is now a mature practice. Foundry (formerly IDG Communications) released the 2023 cloud computing survey, which polled 893 IT decision-makers. In that survey, 66% of respondents expected their cloud spending to rise the following year. Moreover, 34% cited the desire to reduce total cost of ownership, or TCO. However, the question has changed by 2026. It is no longer whether to run a cloud migration. It is how to govern consumption afterward. 

Why companies choose AWS for cloud migration

In practice, companies of all sizes have moved their IT infrastructure to AWS for years. In other words, the stated reason is the cost saving compared with maintaining on-premises environments. Moreover, those savings received independent validation. TechTarget’s Enterprise Strategy Group (ESG) published a report on the topic. According to that study, organizations can cut compute, storage, and networking costs by up to 66%. They achieve this through cloud migration of on-premises workloads to AWS cloud infrastructure. However, the figure comes from the report, not from a contractual guarantee. 

For this reason, it helps to know how the data was built. The ESG report combines qualitative and quantitative economic analysis. For example, it draws on interviews with AWS end users, ESG research, and a conservative economic model. As a result, it illustrates the range of savings an organization can achieve. These come from moving infrastructure to the cloud. In short, the benefit shows up in infrastructure cost, price-performance, efficiency, and user experience. Even so, the real outcome depends on management discipline after go-live. 

Additionally, organizations can build on AWS cloud infrastructure. According to the ESG economic validation report, they benefit from cloud-based innovations. In addition, they reported faster time to value (TTV), greater business agility, and reduced risk. According to ESG, AWS customers maximized savings with monitoring services and tools. Today that work has its own name. It is called FinOps, and it covers chargeback, rightsizing, and usage commitments. Moreover, a new factor entered the math. AI workloads are GPU-intensive and energy-intensive, so they need their own modeling. 

What the TCO model revealed

Likewise, ESG built a conservative 3-year TCO model. It compared using AWS with running the same workloads on traditional on-premises infrastructure. In that model, customers recorded 63% lower compute cost. First, the gain came from Amazon Elastic Compute Cloud (EC2) instances powered by AWS Graviton processors. In addition, they used AWS Lambda for serverless computing. Finally, they applied Amazon EC2 Auto Scaling, Savings Plans, and Amazon EC2 Spot Instances. They also recorded 66% lower networking cost. For example, they eliminated or reduced on-premises network equipment and connectivity. Moreover, services such as AWS Cloud WAN and AWS Direct Connect simplified network operations. 

Finally, the same model showed 69% lower storage costs. For example, customers used Amazon S3 Intelligent-Tiering and Amazon S3 Glacier storage classes. In addition, they adopted Amazon EFS lifecycle management and Amazon EBS SSD Storage. As a result, they replaced traditional on-premises storage arrays. 

Beyond the numbers: real-world cases

The ESG report offers independent validation of moving workloads to AWS Infrastructure as a Service. For example, it cites lower costs, better performance, and greater operational efficiency. However, organizations such as Snap Inc. (Snap) and FORMULA 1 (F1) saw these benefits well before the study. Likewise, Snap cuts storage costs and improves latency to sustain the user experience. After all, Snap builds the popular visual messaging app Snapchat. At the time of the report, it had more than 363 million daily active users. 

The migration nobody noticed

Snapchat started with a focus on ephemeral content, such as photos that disappeared after a few seconds. However, the app became a place to store media and memories for the long term. Snapchatters, as Snapchat users are called, can do so if they wish. As storage needs grew, the company had to optimize cost without slowing performance. Consequently, in 2016, Snap completed its cloud migration to AWS. “We chose to migrate to AWS because of its global reach, excellent performance, and competitive pricing, which in turn gave us the ability to reinvest in our business,” says Vijay Manoharan, manager of Snap’s media delivery platform team. Next, to lower the cost of permanent content, Snap adopted Amazon S3 Glacier Instant Retrieval. In practice, this class delivers low-cost storage for long-lived data. In addition, such data is rarely accessed and still requires millisecond retrieval. As a result, Snap moved more than 2 exabytes of data seamlessly. That volume is roughly the equivalent of 1.5 trillion media files. In short, the shift ran from Amazon S3 Standard-IA to Amazon S3 Glacier Instant Retrieval. 

“The fact that no customer noticed this massive migration to Amazon S3 Glacier Instant Retrieval was a big win for us”