DEV Community

Cover image for Unpopular opinion about "Moving back to on-prem"
Eyal Estrin for AWS Community Builders

Posted on • Originally published at eyal-estrin.Medium

Unpopular opinion about "Moving back to on-prem"

Over the past year, I have seen a lot of posts on social media about organizations moving back from the public cloud to on-prem.

In this blog post, I will explain why I believe it is nothing more than a myth, and why the public cloud is the future.

Introduction

Anyone who follows my posts on social media knows that I am a huge advocate of the public cloud.

Back in 2023, I published a blog post called "How to Avoid Cloud Repatriation", where I explained why I believe that organizations rushed to the public cloud, without having a clear strategy, that would guide them on which workloads are suitable to run in the public cloud, to invest in cost management and employee training, etc.

I am aware of Barclay's report from mid-2024 claiming that according to conversations they had with CIOs, 83% of the surveyed organizations plan to move workloads back to private cloud, while another report from Synergy Research Group (published in August 2024), claiming that "hyper-scale operators are 41% of the worldwide capacity of all data centers", and "Looking ahead to 2029, hyper-scale operators will account for over 60% of all capacity, while on-premises will drop to just 20%".

Analysts claim there is a trend of organizations to move back to on-prem, but the newspapers are far from been filled with customer stories (specifically enterprises), who moved their production workloads from the public cloud to the on-prem.

You may be able to find some stories about small companies (with stable workloads and highly skilled personnel), who decided to move back to on-prem, but it is far from being a trend.

I do not disagree that large workloads in the public cloud will cost an organization a lot of money, but it raises a question:

Has the organization embedded cost as part of any architecture decision from day 1, or has the organization ignored cost for a long time and realized now that the usage of cloud resources costs a lot of money if not managed properly?

Why do I believe the future is in the public cloud?

I am not looking at the public cloud as a solution for all IT questions/issues.

As with any (kind of) new field, an organization must invest in learning the topic from the bottom up, consult with experts, create a cloud strategy, and invest in cost, security, sustainability, and employee training, to be able to get the full benefits of the public cloud.

Let us dig deeper into some of the main areas where we see benefits of the public cloud:

Scalability

One of the huge benefits of the public cloud is the ability to scale horizontally (i.e., add or remove compute, storage, or network resources according to customer demand).

Were you able to horizontally scale using the traditional virtualization on-prem? Yes.

Did you have the capacity to scale virtually unlimited? No. Organizations are always limited by the amount of hardware they purchase and deploy in their on-prem data center.

Data center management

Regardless of what people may believe, most organizations do not have the experience of building and maintaining data centers to be physically secured, energetic sustainable, and to be CSP grade highly available.

Data centers do not produce any business value (unless you are in the data center or hosting industry), and in most cases, moving the responsibility to a cloud provider will be more beneficial for most organizations.

Hardware maintenance

Let us assume your organization decided to purchase expensive hardware for the SAP HANA cluster, or an NVIDIA cluster with the latest GPUs for AI/ML workloads.

In this scenario, your organization will need to pay in advance for several years, train your IT on deployment and maintenance of the purchased hardware (do not forget the cooling of GPUs…), and the moment you complete deploying the new hardware, your organization is in charge of the on-going maintenance, until the hardware will become outdated (probably couple of weeks/months after you purchased the hardware), and now you are stucked with old hardware, that will not be able to suit your business needs (such as the latest GenAI LLMs).

In the public cloud, you pay for the resources that you need, scale as needed, and pay only for the resources being used (unless you decide to go for Spot, or savings plans, to lower the total costs).

Using or experimenting with new technology

In the traditional data center, we are stucked with a static data center mentality, i.e., use what you currently have.

One of the greatest capabilities the public cloud offers us is switching to a dynamic mindset. Business managers would like their organizations to provide new services to their customers, in a short time-to-market.

A new mindset encourages experimentation, allowing development teams to build new products, experiment with them, and if the experiment fails, switch to something else.

One of the examples of experimentation is the spiky usage of GenAI technology. Suddenly everyone is using (or planning to use) LLMs to build solutions (from chatbots, through text summarization, and image or video generation).

Only the public cloud will allow organizations to experiment with the latest hardware and the latest LLMs for building GenAI applications.

If you try to experiment with GenAI, you will have to purchase dedicated hardware (which will soon get outdated and will not be sufficient for your business needs for a long time), and you will suffer from resource limitations (at least when using the latest LLMs).

Storage capacity

In the traditional data center, organizations (almost) always suffer from limited storage capacity.

The more organizations collect data (for business analytics, providing customers added-value, research, AI/ML, etc.), to more data will be produced and needs to be stored.

In the on-prem, you are eventually limited with the amount of storage you can purchase and physically deploy in your data center.

Once organizations (usually large enterprises), store PBs of data in the public cloud, the cost and time to move such amounts of data out of the public cloud to on-prem (or even to another cloud provider), will be so high, that eventually, most organizations will keep their data as is, and it will become a hard decision to move out of their existing cloud provider.

Modern / Cloud-native applications

Building modern applications changes the way organizations develop and deploy new applications.

Most businesses would like to move faster and provide new solutions to their customers.

Although you could develop new applications based on Kubernetes on-prem, the cost and complexity of managing the control plane, and the limited scale capabilities, will make your solution a wannabe cloud. A small and pale version of the public cloud.

You could find Terraform/OpenTofu providers for some of the resources that exist on-prem (mostly for the legacy virtualization), but how do you implement infrastructure-as-code (not to mention policy-as-code) in legacy systems? How will you benefit from automated system deployment capabilities?

Conversation about data residency/data sovereignty

This is a hot topic, at least since the GDPR in the EU became effective in 2018.

Today most public cloud providers have regions in most (if not all) countries with data regulation laws.

Not to mention that 85-90 percent of all IaaS/PaaS solutions are regional, meaning, the CSP will not transfer your data from the EU to the US unless you specifically design your workloads accordingly (due to egress data cost, and service built-in limitations).

If you want to add an extra layer of assurance, choose cloud services that allow you to encrypt your data using customer-managed keys (i.e., keys that the customer controls the key generation and rotation process).

Summary

I am sure we can continue and deep dive into the benefits of the public cloud vs. the limitations of the on-prem data center (or what people sometimes refer to as "private cloud").

For the foreseen future (and I am not saying this as something beneficial), we will continue to see hybrid clouds, while more and more organizations will see the benefits of the public cloud and migrate their production workloads and data to the public cloud.

We will continue to find scenarios where the on-prem and legacy applications will continue to provide value for organizations, but as technology evolves (see GenAI for example), we will see more and more organizations consuming public cloud services.

To gain the full benefit of the public cloud, organizations need to understand how the public cloud can support their business, allowing them to focus on what matters (such as developing new services for their customers), and lower the work on data center maintenance.

Organizations should not neglect cost, security, sustainability, and employee training, to be able to gain the full benefit of the public cloud.

I strongly believe that the public cloud is the future, for developing and innovative solutions, while shipping the hardware and data center responsibility for companies who specialize in this field.

Why do I call it an "unpopular opinion"? When people are reluctant to change, they rather stick with what they know and are familiar with. Change can be challenging, but if organizations embrace the change, look strategically into the future, embed cost into their decisions, and invest in employee training, they will be able to adapt to the change and see its benefits.

About the author

Eyal Estrin is a cloud and information security architect, and the author of the books Cloud Security Handbook and Security for Cloud Native Applications, with more than 20 years in the IT industry.

You can connect with him on social media (https://linktr.ee/eyalestrin).

Opinions are his own and not the views of his employer.

Top comments (0)