Recently, influential individuals and organizations like Puppet, David Hansson, and Basecamp have been sharing their own insights on the cloud. They state that you should consider repatriating infrastructure from the cloud. Most of the conversation centers around the cost of cloud computing. This question is important for your technical, financial, and service teams. In this post, we illustrate why you should instead consider evaluating a hyperscaler like Oracle Cloud Infrastructure (OCI). OCI provides solutions that help you maintain your flexibility and security, while keeping your costs competitive to on premises.

Flexibility and focus

With cloud-based computing, you can shift priorities on the fly with few penalties. Whether you need to scale up for sudden demand, scale down to reduce cost, or experiment with the latest AI technologies to answer the question of your AI strategy, the cloud allows you to deploy the right amount of technology anywhere across the globe. If you’ve already done the work to make your solution flexible and scalable in the cloud, digging deeper to see if repatriation makes sense is worth the effort.

If you depend on consuming the latest technology like GPUs for machine learning (ML), processors, network, and storage technology, using cloud service providers like OCI reduces the risk of obtaining the hardware and personnel required to operate it. Chip shortages and obtaining qualified personnel to operate them are real problems. These issues become less risky through cloud providers. If you decide not to pursue an experiment, you don’t have to worry about repurposing hardware that’s no longer needed. So, you’re free to focus on what’s important: The products and services you bring to the market for your customers.

Security and compliance

Ensuring that your solutions remain available and keep your data safe is an increasingly complicated challenge. Cloud providers like OCI provide standardized security and compliance tools out of the gate. With OCI, the most critical components for security, such as vaults and vulnerability scanning, are included with most key services at no charge.

Protections from Oracle go beyond financial value, and Oracle provides compliance for national, provincial, and local regulations. Oracle’s public cloud was designed to be secure and compliant from the start. As a government user, you don’t pay extra for protections that are built-in to the platform. These services help to simplify and streamline the certification process for regulated workloads. Repatriating production processes in healthcare, finance, and retail can impose significant process hurdles to achieve certification or pass an audit because you’re responsible for the complete set of infrastructure.

Apples versus oranges

Return on investment (ROI), value on investment (VOI), capital expenditures (CAPEX), and operating expenses (OPEX) all boil down to the same question: How much does the solution cost? What opportunities will I have to give up? Are there costs that I’m not considering? Let’s look at a representative company called BoostBuddy. It has approximately $10M in annual revenue, $5M for costs of goods sold (COGS), and $500,000 per year in cloud spending.

Their application enables customers to optimize their business operations in real time. Because of the industry they’re in, they have some performance and uptime service license agreements (SLAs) that require at least 99.95% availability, and a maximum of 50 ms to respond to a request from their primary API. They have a large customer base in the US and roughly 100 TB of data used in production. They rely on managed Kubernetes and Postgres services to deploy their application and have a typical mix of load balancers, storage, and other services, as shown in the following diagram.

Hugging hardware versus cuddling cloud

BoostBuddy needs to find a solution that’s less expensive than $500,000 per year on their current cloud provider. Hardware is increasingly attractive, especially because server hardware is relatively inexpensive these days and seems to be steadily decreasing in price. Even highly redundant configurations with separate disaster recovery sites are within reach of most operations. BoostBuddy is exploring an on-premises solution to replace its current configuration in the cloud with the following parameters:

• Using a colocation data center in the US
• New hardware only
• Capable of handling 95% of highest observed peak workload
• CPU utilization is at 50% daily
• Five-year depreciation schedule

For hardware alone, the amortized rate of approximately $210k is less than their $500k maximum.

Now, let’s compare BoostBuddy’s total CAPEX outlay to a similar, cloud-optimized configuration from OCI. We scale resources automatically when not needed, such as turning down web app resources post-peak and turning up database and analytic resources for generating reports. These practices can result in cloud savings of 50% or more. The following graphic shows what BoostBuddy can expect to have as their architecture on OCI:

At first glance, it appears that OCI is almost twice as expensive as the hardware for running on prem. Why would anyone make this choice? There must be more to the story. Let’s investigate what it takes to move workloads from the cloud to on-premises.

Paying the rent

Cloud providers rent time on their computers. If BoostBuddy is running their own cloud, they must rent rack space or build and maintain their own datacenter. Let’s assume they’re renting rack space at colocation providers.

For BoostBuddy’s footprint, we can expect to spend on rack rent, network bandwidth, remote hands, software licensing, and power. They might also be renting appliances for storage, databases, or network security, but those calculations aren’t included here. Also not included is anything related to human resources. The premise that many have used in cloud repatriation is that they use and retrain their existing personnel.

Let’s compare the following differences now:

On-premises

Hardware:          $209,570.44
Services:           $66,800.00
Rent:              $108,163.40
——————————
Yearly:            $428,441.51

OCI

Infrastructure:    $382,861.39
Licensing:          $32,000.00
——————————
Yearly:            $417,761.39

Excluding potential pitfalls like hypervisor licensing, OS support, and personnel costs, OCI is less expensive on a yearly basis. Why is OCI less expensive? Surely, OCI is just like everyone else, right?

The garden walls

All cloud providers take advantage of enormous scale to get the best cost for hardware and data center spending. You pay for the ability to orchestrate your workloads flexibly on that infrastructure. OCI offers tremendous pricing advantages over our competition for similar compute and storage. You might also find that network charges are contributing to a large portion of your monthly bill. Most cloud providers charge for traffic between zones and egress to the internet. The rate is designed to make you think twice about letting data out.

With OCI, you don’t pay for intraregion network traffic. Network egress is free for the first 10 TB in a month, and then network egress charges are less than a penny per gigabyte. Our example company, BoostBuddy, had roughly 7.5 TB per month of egress, so it didn’t show up on their OCI bill. Let’s assume BoostBuddy experienced rapid growth, and now has 50 TB of internet egress a month. On OCI, that might you an extra $450 per month, but on other clouds, it’s more than a little bit more expensive. In some cases, it’s an order of magnitude more.

Conclusion

Before you consider repatriating to on-premises, you might want to consider OCI. You can maintain all the benefits and flexibility that cloud provides, with price and performance that rivals what it takes to run everything on your own. In our example, BoostBuddy saved approximately $500k compared to staying on their existing cloud provider over 5 years. Even when BoostBuddy needed to grow, OCI’s incremental costs stayed low and prevented surprise egress charges from showing up on the bill.