Migration & Modernization

Unleashing the Power of the Cloud with the AWS Cloud Value Framework (CVF) – Cost Savings (2/7)

Pillar 1: Cost Savings – Total Cost of Ownership (TCO)

Introduction

This blog forms part of a series on the AWS Cloud Value Framework (CVF). The CVF serves as a comprehensive guide to help businesses evaluate, quantify, and communicate the value of AWS Cloud adoption. It comprises five pillars, with this blog focused on Pillar 1: Cost Savings TCO.

In this blog, we will explain:

  • The Cost Savings Pillar
  • The benefits that organizations have been able to achieve
  • How organizations were able to use AWS to achieve these benefits
  • Examples and case studies
  • How to demonstrate the value for your own migration

Pillar 1: Cost Savings TCO

AWS helps organizations reduce their costs. TCO analysis compares the costs of running applications on-premises versus on AWS. The Business Value of Migration to Amazon Web Services, by The Hackett Group found organizations using AWS Cloud achieved the following Cost Saving benefits:

Key performance indicator Overall respondents Top performers
Before migration After migration % change Before migration After migration % change
Total IT spend as % of current annual revenue* 8.15% 7.81% -4% 6.94% 6.08% -12%
Total enterprise technology infrastructure spends as a percent of current annual revenue** 3.71% 2.97% -20% 3.67% 1.95% -47%
Overprovisioning of applications as % of capacity above normal demand 43% 33% -23% 50% 30% -40%

Table 1: Cost savings achieved from migrating to AWS. Table source: Hackett Group, The Business Value of Migration to Amazon Web Services

* Technology spend levels for the study population are not representative of cross-industry averages. This is because of a preponderance of industries that invest more in technology (high tech, financial, retail.) In addition, this is due to a bias in participation by organizations that view technology as highly strategic, and predisposed to greater technology investment.

** The percentage of annual revenue is used to remove the bias of company size in expressing percentage savings and other quantifiable metrics in this study.

To understand the saving, you must understand your current on-premises spend including refresh and equipment replacement.

The image shows a breakdown of the costs that you should consider when comparing an on-premises environment to the cloud. The image shows five categories of costs: 1.Server costs including hardware; servers, rack chassis Power Distribution Units (PDU), Top of Rack (ToR) switches, maintenance, software operating system (OS), virtualization licenses, and maintenance. 2.Storage costs including hardware; storage disks, Storage Area Network (SAN), Fiber Channel (FC) switches, and software backup 3.Network costs including hardware; Local Area Network (LAN) switches, load balancers, bandwidth costs, and software network monitoring 4.Database costs including hardware that is covered in the preceding servers and storage costs, as well as the software database licenses 5.Labor costs including server admin, database admin, storage admin, network admin, support team, project planning, advisors, contractors, managed services, refresh team The image also shows that facilities costs including space, power, and cooling should be considered for servers, storage, network, and databases.

Figure 2: Capturing the costs for your current environment

Figure 2 shows a breakdown of the costs that you should consider when comparing to the cloud. Use this breakdown to itemize your costs. This includes capturing fixed or variable costs, the payment frequency (monthly, yearly), the next refresh date, the cost to refresh, and any End-of-Life Support costs. This will help you build up a view of your spending.

When modeling your spend on AWS, consider building a solution that aligns more closely to your desired future state. For instance, if you migrate an instance to AWS, it may not be fully optimized. We recommend modeling your spend against the optimized state on AWS, rather than its current state. Organizations have been able to optimize their cloud spend using the following levers on AWS:

  • Excess capacity – Are you paying for additional unused capacity on-premises? 451 Research found that on-premises data centers typically have a server utilization rate of under 15%. With the cloud, you only pay for what you use. You can also quickly add (and remove) capacity based on your needs. To help organizations understand their current usage, AWS provides an automated discovery tool, Migration Evaluator. Migration Evaluator provides the insights that you need to build a data-driven business case for migration to AWS, helping define the next steps in your migration journey. It helps you to analyze your current state, define your target state, and develop a migration readiness plan with projected cloud costs to reach your financial objectives. It provides data insights on your current usage and monitors trends to recommend the correct instance sizing and pricing models.
  • Resource utilization – Do you need to run infrastructure 24 x 7? Could you turn off resources during evenings or weekends when developers or testers won’t be working? If the latter, then you could use On-Demand Instance pricing with AWS Instance Scheduler to start and stop Amazon EC2 instances and databases. Suppose you previously had all of your instances running at 100% utilization. This solution can reduce your weekly utilization from 168 to 40 hours, resulting in up to 76% cost savings. Refer to the Instance Scheduler implementation guide for further details.
  • Saving Plans and Reserved Instances – Are you running workloads that need consistent usage? Consider using commitment-based purchasing to optimize your spending, an approach Airbnb has implemented. Savings plans provide savings of up to 72% of your AWS compute usage.
  • Storage access patterns – Do you have a large volume of storage that is infrequently or rarely accessed? It is estimated that by 2025, 80% of all digital data can be classified as archival. This means that organizations are storing large amounts of data that are rarely accessed. When migrating to AWS, review your data storage requirements including your data retention policy and access frequency to determine the appropriate storage class. To remove the complexity of managing storage access, organizations are using Amazon S3 Intelligent-Tiering. Amazon S3 Intelligent-Tiering storage class is designed to optimize storage costs by automatically moving data to the most cost-effective access tier when access patterns change. To manage your objects so that they’re stored cost effectively throughout their lifecycle, create an Amazon S3 Lifecycle configuration. An Amazon S3 Lifecycle configuration is a set of rules that define actions that Amazon S3 applies to a group of objects. If you know your storage access patterns, Amazon S3 Lifecycle configuration will be the most cost-effective solution. Canva is an online design tool that empowers users worldwide to design, edit, and publish anything they can dream up. Canva applied a lifecycle policy to each of their S3 buckets, and quickly migrated nearly 80 billion objects in approximately two days. They then used the Amazon S3 Lifecycle configuration for reducing costs on their infrequently accessed user data. It offers the best of both worlds from S3 Standard-IA and Amazon S3 Glacier Flexible Retrieval. “How Canva saves over $3 million annually in Amazon S3 costs“ blog provides more details about these configurations.
  • Software costs – Is a large portion of your IT spend related to software licenses? We recommend using an AWS Optimization and Licensing Assessment (OLA). The AWS OLA delivers a comprehensive report that models your deployment options. It is based on actual resource use and your existing licensing entitlements, helping you uncover potential cost savings through flexible licensing options. You could also migrate to open source databases like MySQL or PostgreSQL to reduce your software license costs.
  • Cloud-native services – Are you paying for hardware appliances that could be replaced with cloud-native services such as load balancers and network firewalls? When thinking about migrating to native services, there are some considerations:
    • Operating model – when migrating to the cloud, it is recommended you review your operating model. If you attempt the same practices for on-premises environments, it has been shown to limit the potential value achieved with the cloud. Refer to the Building your Cloud Operating Model for initial guidance.
    • Feature set – AWS services are feature rich and used by many large organizations as they address the majority of requirements. However, you may have a complex use case that needs a third-party solution. If you have a complex use case, consider the tradeoffs by simplifying your solution to use cloud-native services. Review the features for AWS Network Firewall, AWS Web Application Firewall, and Elastic Load Balancing.
  • Disaster Recovery – Are you paying for infrastructure, software, and facilities to support your Disaster Recovery (DR) requirements? With the cloud, you can develop DR strategies for each application, where you can balance cost with recovery times by paying for what you use.
  • Fault tolerant workloads – Do you have stateless, fault-tolerant, and flexible applications that can be interrupted? Consider using Amazon Elastic Compute Cloud (Amazon EC2) Spot Instances to take advantage of unused capacity in the AWS Cloud. Spot Instances can help you save up to 90% compared to On-Demand Instance prices. Spot Instances are part of the AWS Well-Architected Framework and recommended purchasing option. They can be used for several workloads including; Amazon Elastic Containers Services (ECS), Amazon Elastic Kubernetes Service (EKS), Amazon EMR, and Amazon Sagemaker. Combine Spot Instances with other purchase models, such as On-Demand Instances and Savings Plans, to maximize cost savings. Also, Spot Instances with Amazon EC2 Auto Scaling to optimize workload cost with performance. Refer to the Amazon EC2 User Guide for best practices and how to use Spot. You can read more on the value creation from Spot Instances in the case study: “amaysim Saves 75% on Compute Costs by Running Amazon EC2 Spot Instances.”
  • Processor choice – Amazon EC2 offers the broadest and deepest compute platform, with over 750 instances. One choice that can help optimize your spend (and improve performance) is the choice of processor. AWS is the first major cloud provider that supports Intel, AMD, and ARM processors. The following table shows an example of the cost benefits that can be achieved. For Graviton (ARM), there is a transition guide that can help you identify compatible workloads and develop your transition plan to Graviton.
Processor choice M7a AMD M7i Intel M7g ARM
.large * $ 0.11592 $ 0.1008 $ 0.0816
% difference between M7a -13% -30%

Table 2: Pricing comparison of different processor option for M7 instances

* Pricing example was taken from 4 May 2024. It uses US East Region, On-Demand Instance Hourly Cost, Linux Operating System, and a Shared Tenancy model.

Demonstrating Business Value

AWS enables organizations, like Kmart Australia, to reduce their infrastructure spend. By migrating to AWS and reducing its hardware footprint, Kmart Australia has saved over $1 million in annual infrastructure, licensing, and other cloud system costs. When developing a TCO, it’s important to understand your current spend and have an accurate baseline as a comparison.

With a baseline of your existing spend, map your future requirements to AWS, and include the preceding commercial levers in your future state modeling. Table 3 provides a sample view of how to list a comparison of your on-premises and AWS costs, including the commercial levers. The numbers in table 3 are for illustration purposes only.

On-Premises * Metric 5 Year Cost AWS * Metric  5 Year Cost
Physical Servers – refresh and maintenance 200 $760,000 Amazon EC2 Instances – AMD Processors 3000
Virtual Servers 2800 $4,310,000
Virtual Hosts – refresh and maintenance 100
Avg. Peak CPU – 40% Avg. Peak RAM – 60% Amazon EC2 Instance rightsized removing excess capacity
Non-production (10 hours x 5 days) 1200 On Demand Instances (30% Uptime) 1200 $1,480,000
Production (24 x 7) 1800 3 Year Saving Plans 1800 $3,110,000
Data Center Racks and Facilities 20 $2,250,000 Included in Amazon EC2 price
Software (OS and Hypervisor) 3000 $1,650,000 Included in Amazon EC2 price
5 Year Compute Total $8,970,000  5 Year Compute Total $4,590,000
Useable On-Premises Storage 660 Tb $3,070,000 AWS Storage (35% SSD, 40% HDD, 25% Object) $2,650,000
5 Year Storage Total $3,070,000 5 Year Storage Total $2,650,000
On-Premises Network (DC Switches, Firewalls, Load Balancers, Data Circuits) $1,486,000 AWS Direct Connects, VPC, Elastic Load Balancing, AWS WAF, Data Transfer, CloudWatch $720,000
5 Year Network Total $1,486,000 5 Year Network Total $720,000
AWS Support $900,000
5 Year AWS Support $900,000
Compute, Storage, and Network Total Over 5 Years $13,526,000 Compute, Storage, Network, and Support Total Over 5 Years $8,860,000
5 Year Cost Savings $4,666,000
% Reduction 34%

Table 3: Example of Compute, Storage, and Network pricing for on-premises and AWS

* Figures are for illustrative purposes only.

You can break this down for each component in your environment. For example: Compute, Storage, Database, Networking, Virtual Desktop, Analytics. Then create a chart (see example in figure 4) and table (see example in table 3) comparing the two scenarios, and calculate the costs savings.

The image is a bar chart showing a 5-year cost comparison between on-premises and AWS using the data from table 3. The chart compares on-premises and AWS costs split by compute, storage, network, AWS Support, and the five-year total. The chart summarizes a potential saving of $4.67 million.

Figure 3: 5-year cost comparison for on-premises and AWS

Additional Support

For additional guidance demonstrating the Business Value, refer to the Cloud Economics Center. Contact your AWS account representative to request a complimentary Cloud Economics assessment.

About the Author