Costs and Benefits of .NET Application Migration to the Cloudv1.0

1. Executive Summary

Thousands of custom applications written using development frameworks such as ASP.NET are deployed by organizations, often running in on-premises environments. At the same time, organizations continue to move larger workloads to the cloud as part of their digital transformation strategies. While the COVID-19 pandemic accelerated these digital transformation initiatives, the competitive advantages realized by organizations have kept the momentum high as business models evolve in the era of AI. The cloud allows them to augment and transform their existing applications using modern developer platforms to increase customer engagement, streamline operations, lower costs, and enhance profitability.

Ripping and replacing large application estates is becoming increasingly difficult for organizations to justify when they are also coping with shrinking budgets. Technology decision-makers must decide on an optimal strategy for these legacy applications, which are often mission-critical to their core business. Companies are often faced with a choice to either retain legacy applications, migrate and modernize to a cloud-based environment, or completely re-architect or replace them. While the third option might appear to be an attractive solution at first glance, for many organizations the overhead in terms of the developer effort, longer time to value, and the cost required to rewrite thousands of applications is untenable.

To help technology decision-makers take a data-driven approach to understand some of these dynamics, we conducted a field test to assess the benefits and costs of application migration. Now in its third year, this benchmark report walks through a migration scenario and evaluates its costs, performance, and benefits. This year’s field test reinforced the following findings:

• Traditional .NET applications can benefit from cloud migration without refactoring underlying code.
• Significant cost savings can be achieved by migrating applications to the cloud from on-premises infrastructures.
• The Microsoft Azure solution offers a potential total cost of ownership (TCO) savings of up to 56% over running on-premises and 39% over running on AWS.
• Streamlined operations, simplified administration, and proximity to advanced cloud services are additional benefits.
• Built-in tools for Visual Studio and MSSQL Manager provide ease of use for database and application migrations to Azure.
• Applications compatible with the migration tools can be simple and straightforward to migrate from VMware to Microsoft Azure. For the application we tested, the Microsoft Azure migration tool successfully migrated the application to the cloud while the AWS migration tool did not.

The field study examines the process of migrating .NET applications and data from on-premises deployment to both Microsoft Azure and Amazon Web Service (AWS) PaaS offerings. Our testing showed that both cloud providers have a strong value proposition for migrating on-premises workloads. Companies can take advantage of Azure Hybrid Benefit, a licensing offer that helps you save up to 85% of the overall cost of SQL workloads in the cloud. This can fundamentally alter the economics behind a data- or transaction-based business model.

Figure 1. Three-Year Comparison of On-Premises, AWS, and Azure Operating Cost in USD

Figure 1 shows that there is a strong case for modernization by implementing auto-scaling and migrating your on-premises workloads to a PaaS solution. This chart depicts optimal minimum pricing, which shows the maximum potential savings on the platform. This may involve specific pricing agreements such as Azure Hybrid Benefit or non-standard architecture such as database translation layers (AWS). The on-premises minimum pricing represents the prorated cost of the application assuming the servers that host it are fully utilized.

In this report, we assess two other deployment scenarios in addition to minimum pricing:

• Baseline is the simplest form of deploying the intended architecture without utilizing cloud-native capabilities such as auto-scaling. It does leverage long-term commitments such as reserved instances.
• Auto-scaling pricing shows the potential savings of deprovisioning unused infrastructure; in this example we run a static workload of 1 node 20 hours per day, and scale up to 4 nodes during a simulated peak traffic of 4 hours per day.

We see a trend line showing Azure leading the pack in terms of environment cost, with the significant savings on SQL Server using Azure Hybrid Benefit delivering a $29,675 TCO over three years. This is a great option for organizations that will continue to operate their on-premises environment. A similarly sized AWS environment would be 20% more expensive than a standalone Azure environment and 67% more expensive than an Azure environment employing Azure Hybrid Benefit licensing, core-for-core.

We will go further into the environment configurations, costs, and price-performance analysis later in the report.

In addition to the licensing cost benefit, we also found that app migration to the cloud reduces the considerable time and resources previously spent on maintaining on-premises resources. There is a positive financial impact when developers and IT staff use that time to work on more strategic tasks that drive value to the business. As such, migration and modernization may be viewed as foundational elements of a forward-thinking enterprise application strategy. We encourage you to leverage application assessment and migration tools, which can evaluate your existing applications estate in bulk and provide recommendations to determine the best migration strategy based on factors such as application readiness and business priorities.

2. Digital Transformation: The Drive for Application Modernization

Digital transformation initiatives have been underway for several years, with many organizations running multi-year strategies, initially accelerated by the COVID-19 pandemic and its economic difficulties. Challenged by reduced budgets, organizations opted to retain more legacy applications, modernize them, and migrate them onto cost-effective cloud platforms.

Thousands of legacy .NET applications are still in use, and rewriting them all is not feasible. While they provide value to the organization, they are often expensive to maintain, inefficient, and cause bottlenecks. Furthermore, application modernization has advantages over developing new applications. Employees are familiar with the applications, and as modernizations improve usability and productivity, they become easier to use and maintain and do not require extensive training.

Modern applications running in the cloud enables organizations to benefit from the security built into the cloud platforms and to focus on their core business, unlike the overhead to continually review and update security measures for on-premises applications. An additional benefit of cloud adoption is that with employees working from home and accessing corporate applications from outside the firewall, organizations can benefit from the cost benefits of the security measures provided by cloud providers. By contrast, securing on-premises systems for remote workers has required additional measures, adding to the overall cost of on-premises environments.

3. Options and Benefits for Application Modernization

Application modernization runs the spectrum from a complete refactor and rewrite of the application to a simple lift-and-shift of the application to a new platform. Terminology may vary, but we can focus on the following options, depicted in Table 1.

Rehosting is the simplest method, often referred to as a “Lift and Shift.” This involves simply moving your VMs from one hosting provider to another, such as your on-premises data center to AWS or Azure. While this is the lowest-cost option, it also gives the least opportunity for modernizing and reducing technical debt.

Replatforming involves changing the underlying platform that hosts your application, such as moving your VM-based SQL Server workloads to a fully managed cloud-based SQL provider such as Azure SQL or Amazon RDS. This could also refer to moving a VM-based container workload to a containerization service such as Azure Kubernetes Service or Amazon EKS.

Repurchase would refer to replacing a self-hosted application with a SaaS product, such as replacing your on-premises version control system, like GitLab or Team Foundation Server, with a cloud-based solution such as Azure DevOps or GitHub.

Refactor/Re-architect is generally the most expensive option and involves a rewrite of the applications to run natively on a new platform. This re-architecture could be paired with a move to serverless architecture using Azure Functions or AWS Lambda functions or rebuilding a monolithic application into containerized microservices behind a service bus.

Retire is an often underutilized approach as organizations are reluctant to change processes and reduce technical debt. An excellent example of software to retire would be infrastructure monitoring systems for on-premises environments. These simply don’t need to be migrated to cloud, as this functionality is already available out-of-the-box from the cloud platforms.

Table 1. Benefits and Challenges of Modernization Approaches

Table 1 shows that a replatform can provide a happy medium as it removes the need to maintain the infrastructure, which is still required with a lift-and-shift approach, but it is not as time-consuming, costly or complex as a rewrite. With replatforming, companies benefit from having the cloud provider manage the underlying infrastructure and software of the platform while only needing to make minor changes to their applications.

Alongside potential cost reductions, the benefits of replatforming include:

• Increase developer productivity
• Enhance use of cloud features
• Focus on application innovation

These benefits respond to the need for greater agility to support digital transformation. Each is covered below, with illustrations from the Microsoft Azure portfolio.

Increase Developer Productivity

Migrating an application from an on-premises deployment to a fully managed service in the public cloud can increase productivity in the following distinct ways:

• Reduce administrative overhead: Adopting a managed platform helps optimize costs and lowers administrative overhead associated with the application. The operations team no longer needs to patch operating systems, manually scale for capacity, or manage underlying database engines.
• Improve operational efficiency: Reduced overhead means the operations team is freed up to focus on other aspects of the system, working to improve operational efficiency and reduce downtime. This moves operations onto the front foot, enabling teams to operate more confidently.
• Speed application and feature delivery: Rather than just keeping the lights on, the ultimate goal of operations is to act as an enabler of innovation, empowering development teams to deliver new applications and features faster by reducing deployment times and enabling access to leading-edge services.

Providers like Microsoft Azure offer rich monitoring, reporting, and analytical tools that will assist the operations team in diagnosing potential issues, finding the root cause, and verifying that a solution is effective. Both Azure App Service and Azure SQL Database scale dynamically and automatically to meet customer needs and application demand. Using these capabilities, operations teams no longer have to spend time manually spinning up new instances or vertically scaling their database server to keep pace with demand. They also do not need to plan for capacity in their data centers to accommodate peak load and growth. The cloud can scale to meet bursts of activity and then contract when demand subsides.

Cloud providers can also automate the provisioning of additional operational environments for development, quality assurance, staging, and more. The operations team no longer needs to spend weeks creating an additional environment, and they can use infrastructure as code to ensure the various operational environments stay consistent.

Enhance Use of Cloud Features

Beyond operational efficiency, cloud providers offer many features that enhance both the operational and development experience. For example, the Azure App Service platform enables developers to use deployment slots on a web application. Each deployment slot runs its version of the application with an independent namespace. Traffic can be distributed across multiple slots for canary deployments, blue/green testing, or a regular cutover between staging and production. Because there is also deep integration between Visual Studio and Azure App Service, additional capabilities, such as live-site debugging, are unlocked for the developer.

Azure App Service also integrates with GitHub, Azure Repos, and BitBucket to enable continuous deployment. New commits to a specific branch in a code repository will trigger a build process and deployment of the new code to Azure App Service. The continuous deployment feature can be integrated with Azure Pipelines to ensure proper code coverage and testing are performed before deploying a new build.

Backup and recovery of application data and replication to a separate region are simplified when using the cloud. Azure SQL Database can easily replicate data to a designated partner region. A scaled-down instance of Azure App Service could be running in the same partner region, with the whole solution placed behind a Traffic Manager to make failover a relatively simple process. Since capacity can be allocated on demand, it is no longer necessary to maintain a disaster recovery environment with matching capacity at all times.

Focus on App Innovation

Migrating on-premises applications places them directly adjacent to the rapid innovation happening in the cloud. New services and features are constantly introduced in clouds like Microsoft Azure. In some cases, an application might be updated to use a different service, like moving from traditional Microsoft SQL Server to Azure SQL Database to handle frequent scaling and unpredictable use. In other cases, an application could be enhanced by consuming another service, like adding a chatbot to an existing application using Azure Bot Service or adding voice commands with Azure Cognitive Services.

By moving an application to the cloud, it will have direct access to all of these services and more. Additionally, developers and operations managers are freed up to investigate and experiment with these services. Instead of spending cycles on routine infrastructure management tasks, precious developer resources can be directed to application innovation.

4. Comparing PaaS Offerings For Migration

While newer applications might be developed using either platform as a service or in a cloud-native manner, traditional ASP.NET applications were written before advances like .NET Core and microservice architectures. It is not simple or cost-effective to rewrite these applications to support a new programming paradigm, but they can still take advantage of what the cloud offers.

Traditional ASP.NET applications backed by Microsoft SQL Server can be moved to the PaaS offerings on the major public cloud providers. Microsoft offers Azure App Service and Azure SQL Database to provide the web front end and database back end required by most applications. As a managed service, Microsoft takes care of the underlying infrastructure management and configuration, reducing the operational overhead and cost associated with managing and maintaining web and database servers.

AWS has similar capabilities in its Elastic Beanstalk service, which can generate environments that include an autoscaling group of EC2 instances and a deployment of its Relational Database Service. Both AWS and Microsoft provide managed virtual machines running Windows Server and IIS, mirroring the current deployment environment of traditional ASP.NET applications. Microsoft’s PaaS solution abstracts most of the underlying infrastructure and maintenance details, while AWS’ offering exposes these to the customer.

The similarities of these platforms make the migration of ASP.NET applications from on-premises environments simple and predictable while providing benefits over a lift-and-shift migration that simply uses infrastructure-as-a-service (IaaS) components.

Essentially, traditional applications can be moved seamlessly to a PaaS deployment in the cloud without refactoring or rewriting code.

We selected OrchardCMS as our benchmark application, and we conducted a number of tests related to migrating this application to both Azure App Service and AWS Elastic Beanstalk.

To simulate the on-premises environment, the .NET application is deployed to virtual machines on a VMware host. The application was then migrated to PaaS environments, using automated tools from Azure and Amazon Web Service, respectively.

We focus our findings on the following elements:

• Migration Process: We look at the complexity, available tooling, and issues encountered.
• Price-Performance Comparison: We compare the relative performance and cost of each solution to establish best value.
• TCO Analysis: We determine a three-year estimated cost of operation, including platform, licensing, and operational (people) costs.

5. Field Test Results and TCO Analysis

Migration Process

Azure has a slight advantage when it comes to migration tools for .NET and Microsoft SQL Server. In our experience, the AWS database migration service only handles heterogeneous migrations, and the migration assistant for Elastic Beanstalk did not work for us out of the box.

Azure also stood out in terms of ease of use and organization of documentation. While this is a subjective measure, finding documentation for migration in AWS involves searching separate docs for Elastic Beanstalk and the underlying AWS service. Azure had a more centralized approach.

TCO Analysis
Calculating a true TCO for an application modernization or migration effort is challenging because of variables around things like administrative overhead and operational headcount. For the purpose of this analysis, we considered the direct cost of running each environment for three years and made the assumption that headcount remains consistent. We assume that this project is standalone, not part of any broader data center migration or cloud deployment projects.

Hosting considerations also need to be accounted for. Migrating the entire application from an on-premises environment would likely result in the removal of one virtualization host. Therefore, for the on-premises environment, we included the cost of two physical hosts, four vSphere licenses for the CPUs, and an estimate for the power and cooling consumed by those hosts.

In terms of results, an Azure configuration with Azure Hybrid Benefit applied had the lowest overall cost, saving an estimated 56% over the on-premises deployment and 39% over AWS.

The Azure Hybrid Benefit licensing offer allows existing Windows Server and SQL Server licenses with active Software Assurance to be applied to Azure virtual machines and Azure SQL Database instances. The estimated cost of a SQL Standard license is $13,595.94, based on CDW published in August 2023. Assuming a migration scenario with an existing SQL Standard License being recouped for use in Microsoft Azure, the Azure SQL Database cost would be reduced from $529.27 to $237.37, which would reduce the total Microsoft Azure cost to $32,226. The use of existing SQL licenses is not an option on Amazon RDS. The use of Azure Hybrid Benefit creates a compelling savings of 39% over the overall AWS environment and a database-specific savings of 71% over RDS for SQL Server.

These relative costs are shown in Table 2. Note that we do not incorporate dynamic scalability and bundled features, which also have a bearing. In addition, the pricing calculation for Azure uses reserved instance pricing for App Service and Azure SQL and the East US region for hosting, while AWS uses reserved instance pricing and the us-east-1 region for hosting. More information regarding the configuration and calculation can be found in the appendix.

Table 2. Estimated Cost by Environment

There are other, less quantitative aspects that we believe give Microsoft Azure an edge when calculating a true TCO figure. One important aspect is operational efficiency and administrative burden. Each approach offers a different combination of benefits and costs:

• Azure App Service and Azure SQL Database are designed to be true, managed platforms, abstracting the underlying components that compose each service. To a certain extent, this limits the amount of customization and control that a customer has over the web service or database. The tradeoff is that Microsoft handles upgrades, patching, and maintenance.
• Amazon Elastic Beanstalk acts more like an automation platform to provision an environment than a fully managed offering. The base constructs used by Elastic Beanstalk—EC2, RDS, ALB—are not abstracted from the customer, providing a large range of flexibility. The tradeoff for AWS customers is a potential for a greater administrative load that is more akin to IaaS. In fact, a more apt comparison might be Azure Virtual Machine Scale Sets, Application Gateway, and Azure SQL Managed Instances.
• On-premises deployment does not offer any form of managed services. It is up to the existing operations teams to continue to support the application. Any migration to the cloud should be considered in terms of the additional administrative load and complexity introduced by a new environment, compared to the lower cost and operational efficiency of using a managed service.

Finally, we acknowledge the additional flexibility provided by an IaaS solution may be required by some applications. The test applications used in our benchmarking represent enterprise web apps that benefit from PaaS. Based on Microsoft’s investments and focus on supporting traditional ASP.NET, most applications should work on Azure App Service.

Price-Performance Comparison

To calculate the price-performance of each approach, we took the three-year TCO of each environment and divided it by the login throughput achieved in each environment. The resulting price-performance figures (shown in Figure 2) revealed the following:

• The Azure environment with Azure Hybrid Benefit applied achieved the lowest cost factor overall.
• Both cloud environments, when employing auto-scaling, achieved significant value advantages over on-premises

Figure 2. Price-Performance Based on Login Test Throughput (USD)

Regardless of whether the application is running on-premises or on Microsoft Azure, the same SQL Standard licenses would be necessary. If the application is moved to AWS, the license cannot be migrated, and instead, the license cost is rolled into RDS. Therefore, the Microsoft Azure environment with Azure Hybrid Benefit represents a performance-normalized cost savings of 16% over AWS and a savings of 58% over on-premises VMware.

6. Conclusion

Organizations looking to transform can see application migration as a clear opportunity to reduce on-premises legacy workloads and benefit from the advantages of cloud-based models. As illustrated by this field test for .NET applications, migration to a cloud platform can offer significant benefits over on-premises solutions, both in tangible (cost/performance) and less tangible (access to innovation) terms.

We would recommend that organizations looking to modernize applications in general, and .NET applications in particular, consider the following:

• Existing legacy and heritage applications can be run at a significantly lower cost using a cloud-based platform, freeing resources and unlocking the potential for new application delivery.
• Replatforming applications are minimally disruptive and can be simple to test, enabling piloted migration, risk assessment, and cost evaluation before full deployment.
• Fully managed platforms provide the highest simplicity and the lowest administrative burden compared to in-house infrastructure and platform stacks.
• Migrating applications to the cloud provides access to additional capabilities offered by the cloud providers.
• For .NET applications, it is easy to get started right away by taking advantage of Microsoft’s App Service and Database Migration Assistance tools, which help automate migration to the Azure App Service and Azure SQL Database.

To get started down a migration path, we recommend taking the following actions:

• Assessment: Categorize existing applications by revenue generation and cost to maintain. While there may be hundreds or thousands of applications to treat, consider working on applications according to a specific group, such as department, application, or data type.
• Prioritization: Identify applications that make good migration candidates according to criteria such as existing maintenance cost, migration risk, and opportunity created to unlock new functionality.
• Evaluation: You can map application dependencies across data sources, services, and infrastructure to help identify group migration candidates.
• Validation: Perform test migrations to validate functionality and performance and build experience around target environments, for example, in terms of what operational management dashboards are available.

Keep in mind that moving applications to the cloud does not have to be an all-or-nothing deal that requires either moving existing virtual machines partially or completely rewriting applications. By considering replatforming as an approach, organizations can derive benefits rapidly without cost and risk and unlock new opportunities for innovation and transformation.

7. Appendix

Environment Configuration

On-Premises Configuration

The on-premises environment runs on a single, bare-metal VMWare virtual host located in an Equinix data center in Los Angeles, California. Four VMs are deployed as web servers, one as a database and one as a load balancer. Each web and database VM is assigned four vCPUs, 16 GB RAM, and a 250 GB virtual drive on NVME storage.

Virtual Host Configuration

Processor: AMD EPYC 7502P 32-Core Processor @ 2.50 GHz

Memory: 256GB

Storage: 2x 3.8TB NVME SSD (guest), 2x 240GB SSD (hypervisor)

Networking: 2x 25Gb/s

OS: VMware ESXi 7.0U2a

VM Configuration

vCPUs: 4

Memory: 16GB

Storage: 256GB

OS: Windows Server 2022

AWS Configuration

The AWS environment was deployed in the us-west-2 region (Oregon) in a single availability zone. The latest M7i general purpose instance type was used.

Instance Type: m7i.xlarge

Processor: Intel Xeon Scalable (Sapphire Rapids)

vCPUs: 4

Memory: 16GB

Storage: 256GB EBS (gp2)

Networking: Up to 12.5 Gb/s

OS: Windows Server 2022

In AWS, the database was hosted using Amazon RDS for SQL Server, and the latest instance class is db.m6i.xlarge.

Azure App Service Configuration

The Azure App Service environment was deployed in the west-us-3 (Washington) region in a single availability zone. The App was deployed using the Premium v3 P2V3 SKU (equivalent to a Standard_D4d_v4 virtual machine) with 4 vCPU and 16GB RAM. The app was configured to scale up to 4 nodes.

App Service SKU: Premium v3 P2V3

vCPUs: 4

Memory: 16GB

Storage: 256GB Standard SSD

Networking: Up to 12.5 Gb/s

OS: Windows Server 2022

Load Testing

Load testing was conducted using Azure Load Testing from the South Central US (Texas) region to normalize latency to the west-coast workload locations.

As described earlier, two types of tests were conducted at 500 users and 1,000 users, respectively: login tests using jMeter and page load tests.

TCO Detail

On-Premises TCO Calculation

Table 3 shows the total cost of ownership over three years. This pricing does not reflect the ancillary costs of a true enterprise on-premises environment, which would include networking appliances and support, enterprise storage arrays and support, and management and backup software. The cost displayed here is only the specific cost for adding this application to an existing data center. Depending on the scale of the environment, there could be $10,000s to $100,000s in additional costs associated with an on-premises environment, which are not accounted for here but would need to be pro-rated into the TCO based on this application’s percentage of the total workload of the organization.

Table 3. On-Premises TCO

Here is how we determined costs for the on-premises environment:

• Windows Server 2022 Licensing calculated using 1 physical processor, 32 cores, and 5 VMs. CALs and more nuanced Microsoft licensing line items will vary by organization. Licensing Cost via CDW: https://www.cdw.com/product/microsoft-windows-server-2022-datacenter-license-16-cores/6688190?pfm=srh
• MS SQL Server Licensing is either per-core or per-server + CAL. We are calculating pricing using cores. License cost via CDW: https://www.cdw.com/product/microsoft-sql-server-standard-core-edition-license-software-assurance/2669826
• vSphere Enterprise Plus 8 Licensing is significantly simpler than earlier versions for our single-host setup. License cost via VMWare: https://store-us.vmware.com/vmware-vsphere-enterprise-plus-5653277800.html
• On-Premises Hardware pricing is calculated via Equinix Bare Metal on a 3-year commitment. https://deploy.equinix.com/pricing/

Additionally, as this deployment only leverages 24 of the 32 available cores on the host, the hardware and licensing cost for this application has been prorated by 75%, assuming the remainder of the host would be fully provisioned.

Now, let’s look at the TCO for both AWS and Azure. Note that as cloud providers charge for network traffic egress to the internet, 1 TB per month has been factored into the pricing for both AWS and Azure. On the AWS pricing calculator, these costs are built into the EC2 Instances. On Azure, they will appear as a separate line item.

AWS TCO Calculation

Table 4 shows the base cost of the AWS configuration we used for testing.

Table 4. Baseline Cost for AWS

Baseline AWS Configuration

Pricing Calculator: Link

Bandwidth: 1 TB/month

Auto-Scaling: Disabled

Term: 3-year reserved, no upfront

Licenses: Included

The pricing example in Table 4 is the baseline pricing of a 4-node AWS environment before implementing auto-scaling. This environment would be operating at 100% capacity at all times.

https://calculator.aws/#/estimate?id=0c2462927931e21aac37af0d9bd2588f4c23a51b

Table 5. AWS Pricing with Auto-Scaling

Auto-Scaling AWS Configuration

Pricing Calculator: Link

Bandwidth: 1 TB/month

Scale Sets: 1 reserved node; 3 on-demand nodes

Term: 3-year reserved, no upfront

Licenses: Included

The pricing in Table 5 reflects an auto-scaling group that scales from 1 to 4 nodes for a simulated high-traffic period of 4 hours per day. While real-world traffic may not be this predictable, this scaling ratio will be used across both cloud environments to compare costs.

Most companies do not operate at 100% load at all times, but instead have periods of high and low traffic. This example shows an environment that scales from 1 to 4 nodes during a daily 4-hour window of high traffic, which realizes a 53% reduction in compute costs and nearly 30% overall cost savings. This is a key strategy for cloud computing, and this is the example TCO we use for AWS.

Table 6. Reducing Cost with Babelfish for Aurora

Babelfish for Aurora AWS Configuration

Pricing Calculator: Link

Bandwidth: 1 TB/month

Scale Sets: 1 reserved node; 3 on-demand nodes; Babelfish for Aurora

Term: 3-year reserved, no upfront

Licenses: Included

As Table 6 shows, AWS does have further options to reduce costs while migrating to the cloud: Babelfish for Aurora PostgreSQL. However, this is highly dependent on whether your application works with the translation layer and whether you have the human capital to invest not only in migrating your database, but converting it from SQL to PostgreSQL and validating its functionality. https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/babelfish-migration.html

If this solution works for your application, the RDS costs would be reduced significantly. Due to the large number of dependent factors, it would be impossible to estimate the development and testing time and/or the likelihood this solution is compatible with most enterprise applications running on Microsoft SQL Server. As such, we have chosen to use the RDS for SQL Server pricing for this benchmark. Further savings could be achieved by using flex-instance classes, but again, this is highly dependent on workload and the frequency of high-traffic periods.

Azure TCO Calculation

In contrast to AWS Elastic Beanstalk, which is merely a management layer for the same infrastructure you would deploy using EC2 and RDS, Azure App Service abstracts the compute, load balancer, and storage into a single deployable resource.

We apply the same pricing methods here as we did for AWS. Table 7 shows the baseline cost, and Table 8 shows load-balanced cost.

Table 7. Baseline Cost for Azure

Baseline Azure Configuration

Pricing Calculator: Link

Scale-Out: Disabled

Term: 3-year reserved, no upfront

Licenses: Included

Table 8. Cost of Azure with Auto-Scaling

Auto-Scaling Azure Configuration

Pricing Calculator: Link

Scale Sets: 1 reserved node; 3 on-demand

Term: 3-year reserved, no upfront

Licenses: Included

When customers apply Azure Hybrid Benefit in their subscription, it enables savings up to 85% off of pay-as-you-go pricing for Windows Server and Microsoft SQL Server workloads. https://azure.microsoft.com/en-us/pricing/hybrid-benefit/#why-azure-hybrid-benefit

Table 9 shows the advantages of this licensing option.

Table 9. Reduced Licensing Cost with Azure Hybrid Benefit

Azure Hybrid Benefit Azure Configuration

Pricing Calculator: Link

Scale Sets: 1 reserved node; 3 on-demand

Term: 3-year reserved, no upfront

Licenses: Azure Hybrid Benefit

Azure Hybrid Benefit licensing offers compelling pricing; however, it is available only to organizations with active Software Assurance subscriptions for on-premises cores. Once you successfully exit your data center and no longer pay for on-premises licensing, the additional costs would return to the Azure resources. In cases where companies plan to maintain long-term on-premises and cloud-based production workloads, the Azure Hybrid Benefit is without equal.

Performance Analysis

To determine which solution provides the best value, we performed two kinds of tests on the migrated application: page load testing and login testing. All auto-scaling groups were statically configured to four nodes for the performance tests.

Page Load Testing

We simulated loads of 500 and 1,000 users repeatedly loading the home page of our web application and, as shown in Figures 3 and 4, verified that web server performance scales as a function of CPU performance and disk speed.

Figure 3. Page Load Throughput – 90th Percentile (Page Loads per Second)

Figure 4. Page Load Response Time – 90th Percentile (in Milliseconds)

Figure 4 shows that the 1000-user synthetic workload fully consumed the available performance on the web servers, resulting in 500ms and above response times.

Login Testing

We simulated loads of 500 and 1,000 users repeatedly logging into our web application to test application and database performance. Figures 5 and 6 show the results.

We saw again that CPU class heavily influences performance, but with a database component to the test, the performance gap narrows. We also noted that more recent CPUs have a noticeable throughput difference, with the AWS M7i class instances leading the pack.

Figure 5. Login Test Throughput – 90th Percentile (Logins Per Second)

Figure 6. Login Response Time – 90th Percentile (in Milliseconds)

In performance testing, the 90th percentile measure indicates the time below which 90% of transactions complete. Quite consistently, we found that Azure processes a lower transaction volume but also has a lower 90th percentile login test run time than AWS, resulting in the web server outperforming Azure but the database underperforming. In other words, we saw overall higher throughput from AWS, but the top 10% slowest requests took longer on AWS than on Azure.

8. About William McKnight

William McKnight is a former Fortune 50 technology executive and database engineer. An Ernst & Young Entrepreneur of the Year finalist and frequent best practices judge, he helps enterprise clients with action plans, architectures, strategies, and technology tools to manage information.

Currently, William is an analyst for GigaOm Research who takes corporate information and turns it into a bottom-line-enhancing asset. He has worked with Dong Energy, France Telecom, Pfizer, Samba Bank, ScotiaBank, Teva Pharmaceuticals, and Verizon, among many others. William focuses on delivering business value and solving business problems utilizing proven approaches in information management.

9. About GigaOm

GigaOm provides technical, operational, and business advice for IT’s strategic digital enterprise and business initiatives. Enterprise business leaders, CIOs, and technology organizations partner with GigaOm for practical, actionable, strategic, and visionary advice for modernizing and transforming their business. GigaOm’s advice empowers enterprises to successfully compete in an increasingly complicated business atmosphere that requires a solid understanding of constantly changing customer demands.

GigaOm works directly with enterprises both inside and outside of the IT organization to apply proven research and methodologies designed to avoid pitfalls and roadblocks while balancing risk and innovation. Research methodologies include but are not limited to adoption and benchmarking surveys, use cases, interviews, ROI/TCO, market landscapes, strategic trends, and technical benchmarks. Our analysts possess 20+ years of experience advising a spectrum of clients from early adopters to mainstream enterprises.

GigaOm’s perspective is that of the unbiased enterprise practitioner. Through this perspective, GigaOm connects with engaged and loyal subscribers on a deep and meaningful level.

10. Copyright

© Knowingly, Inc. 2024 "Costs and Benefits of .NET Application Migration to the Cloud" is a trademark of Knowingly, Inc. For permission to reproduce this report, please contact sales@gigaom.com.