AWS Cloud Financial Management

Five things you should do to create an accurate on premises vs cloud comparison model

Voiced by Amazon Polly

For CEOs, cloud adoption is not just an engine for revenue growth and efficiency. Its speed, scale, innovation, and productivity benefits are essential to the pursuit of broader digital business opportunities, now and well into the future.” – from McKinsey’s report, Cloud’s trillion-dollar prize is up for grabs

Organizations are adopting AWS as a differentiator to be more agile, innovative, and efficient, and cost savings is often used as a powerful conversation starter when justifying the move to internal stakeholders.

Numerous independent and AWS-led studies have demonstrated the cost savings organizations achieve by migrating to the cloud. For instance, an IDC study estimated that AWS customers have a 51% lower cost of operations compared to running on premises infrastructure. Results like this have been re-affirmed by thousands of AWS customer testimonials.  Some notable customer stories include: GE, who reduced its Total Cost of Ownership (TCO) by more than 50%, and Dow Jones, who estimated global infrastructure cost savings of $100 million by migrating its data centers to AWS.

When helping customers build TCO models for their cloud migration, we often come across customer analysts who say their financial models show that moving to the cloud is more expensive than staying on premises. What we’ve found is that customers often exclude key inputs that lead to inaccurate cost comparison models.

How to build an effective cost comparison model

Here are five key actions you can take to help create an effective cost comparison model:

1.  Include all on premises cost components

While calculating the on premises costs, you should make sure you’re looking at all the cost components that go into building and maintaining your data centers. These costs can range from upfront capital to regular operational expenditures that keep the data center running. Even though capital expenditures such as hardware, racks, and network equipment are a one-time purchase, they usually have a refresh cycle of five years. You should amortize this expense for comparison. The figure below shows some key on premises components.

on-premises cost model considerations

Figure 1. Components to consider for on premises cost

2.  Understand that capacity forecasting for cloud is different than on premises

Capacity planning can be a major challenge for a traditional data center. When predicting demand and building capacity for an on premises environment, you need to consider the spikes in demand curve. It requires accurately predicting technology needs and procuring equipment, typically six to nine months in advance. Customers have to overprovision anywhere between 20% to 50% over peak-predicted demand to meet the spiky workloads. This overcapacity leads to a lot of underutilized resources.

The National Resources Defense Council finds that average server utilizations can range between 12% to 18%, and a recent McKinsey study found that over 30% of servers were utilized less than 10% of the time. Those industry averages indicate exactly why there is so much value to be realized from migrating to the cloud.  Cloud can provide the elasticity to spin up or spin down resources based on need, thereby reducing overhead, as depicted in the model below.

chart showing capacity planning difference for on-prem vs AWS

Figure 2. Shows difference in capacity planning for on premises vs AWS

Mapping to the correct cloud instance type needs to be based on actual utilization, rather than exact mapping to provisioned on premises servers. For example, a server with 12 cores and 64 GB RAM on premises having CPU utilization of 25% and RAM utilization of 25% can be easily mapped to a 4 vCPU and 16 GB m5.xlarge or a t3.xlarge instance type depending on the use case. The realities of the on-prem operating model need to be included in an organization’s cost forecast to be able to compare the true benefits of moving to cloud. We suggest running a free Migration Evaluator assessment to evaluate your actual utilization numbers.

3.  Use the optimization options available for existing licenses

Without optimizing your licensing in cloud migration, the cost of overprovisioning third-party licensing can exceed the cost of compute. Windows licensing can sometimes cost 50% or more of compute cost. The cost is even greater if SQL or Oracle licenses are involved. Licensing costs create efficiency opportunities on cloud. Existing low-cost licensing agreements can incentivize bringing your own license (BYOL) and opportunities for significant savings. However, for instances with low uptime, using license-included machines allows an organization to minimize these licensing costs and eliminate the need to procure and manage additional licenses.

The AWS Optimization and Licensing Assessment (OLA) helps customers assess and optimize their licensing cost as they plan their cloud migration. Use its recommendations to get the most value from your existing licensing entitlements. You’ll be able to configure your instances to require fewer licenses while still maintaining high-performing applications.

The complexities of your licensing agreements, the favorable licensing options available on cloud, and the corresponding licensing cost avoidance need to be considered and built into an accurate cost comparison between on premises and cloud.

4.  Choose the right compute pricing model

AWS offers Savings Plans as a flexible pricing model with lower prices compared to on-demand pricing (up to 72% lower in some cases), in exchange for a specific usage commitment (measured in $/hour) over a one- or three-year period. Despite this, some organizations only consider on-demand pricing in their financial models. While on-demand pricing has its use cases, this pricing model should only be used for important but irregular, infrequent, and low-to-moderate uptime workloads (less than 50% daily hours).

AWS also provides “Spot” as a pricing model, allowing organizations to run workloads on unused Amazon Elastic Compute Cloud (Amazon EC2) capacity, which can provide up to a 90% discount from on-demand pricing without making a term-based commitment.

5.  Select the right storage class

Carefully selecting the right storage class based on specific data needs is crucial for you to determine the correct TCO while migrating storage to cloud. High performance block and file storage can be expensive and should be used for data that requires higher input/output operations per second (IOPs). Documents, pictures, and objects can be mapped to inexpensive storage tiers within Amazon Simple Storage Service (Amazon S3). Amazon S3 offers multiple storage classes and S3 Intelligent-Tiering is an ideal storage class for data with unknown, ever-changing, or unpredictable access patterns, independent of object size or retention period. You can use S3 Intelligent-Tiering as the default storage class for virtually any workload, especially data lakes, data analytics, new applications, and user-generated content.

In fact, AWS offers more than 15 different storage categories to meet customer workload needs.

For Disaster recovery needs, AWS provides resources to enable a multi-region or a multi-AZ approach for your workload depending on your Recovery Time Objectives (RTO) and Recovery Point Objectives (RPO). This provides assurance against events that can impact multiple data centers across separate and distinct locations.

Simply configure the Disaster Recovery (DR) and activate it when needed. You won’t incur additional compute cost until it’s activated. Also, S3 data is backed up across multiple availability zones in the cloud, minimizing the risk of lost data in a DR scenario. Your true cost comparison model needs to reflect this advantage.

The big picture on cloud value

Now that we’ve covered the ways you can improve your on premises vs. cloud cost models, it’s important that you take a step back and understand the overarching achievable business value of migrating to the cloud. Because these distinct advantages (cost savings, staff productivity, operational resiliency, business agility) may be more difficult to recognize and quantify before migrating, it’s important to discuss how your organization can define each of them throughout its cloud migration journey. Doing so will help your team recognize and realize the opportunities that the cloud can create, accelerating your organization’s efficiency and innovation.

Contact the AWS Cloud Economics team with your requests, or align with your AWS account managers for help regarding business cases or modeling your AWS costs.

Related info

https://aws.amazon.com/about-aws/global-infrastructure/

https://aws.amazon.com/blogs/aws-cost-management/amazon-ec2-15th-years-of-optimizing-and-saving-your-it-costs/

https://aws.amazon.com/blogs/aws/ec2-price-reduction-for-ec2-instance-saving-plans-and-standard-reserved-instances/ 

https://aws.amazon.com/aws-cost-management/strategic-IT-planning-and-evaluation/

Ankit Malhotra

Ankit Malhotra

Ankit is a Business Value Consultant with the US Public Sector Cloud Economics team and is based in California. He advises customers throughout the public sector and commercial space on quantifying their business value while moving from on-premises to AWS, and builds business cases and TCO models to help customers quantify their savings from the cloud. Ankit holds an MBA and AWS Associate SA certification.

Morteza Zijerdi

Morteza Zijerdi

Morteza leads the US Public Sector Cloud Economics team and is based in Arlington, Virginia. His background is in consulting, finance, and banking. At AWS, he has developed 200+ business cases and been involved in 40+ enterprise migrations for customers ranging from healthcare, defense, federal, SLG, and education. He holds an MBA, and is a CFA Charterholder and a CPA.