|

Why Companies Are Moving Back to On-Premise Software: The Cloud Repatriation Trend

Why Companies Are Moving Back to On-Premise Software: The Cloud Repatriation Trend

Introduction: The Cloud Isn’t Always the Promised Land

Do you recall all that technology jamboree that shouted out that you have to migrate to the cloud or die? Let me put it this way, real life has a tendency to bring even the most bombastic trains of hype down the notches. An increasing number of companies, both greenhorn startups and major Fortune 500 corporations are quietly reversing load in the clouds. Why? Due to the fact that the sparkling promises made by the cloud are usually accompanied by sets of costs to pay, headaches due to compliance, and compromises in terms of performance.

Consider 37Signals, which creates Basecamp and HEY. Their costly abandonment of AWS, abandoning a million dollars a year of cloud costs in favor of reengagement with on-prem servers, was legendary. Or look at European banks who are now regretting their public cloud use because of hard rules of GDPR. The truth? Cloud repatriation is not a marginal sect, it is a business turnaround.

The Cloud’s Dirty Little Secret: It’s Expensive (Really Expensive)

It was thought that the cloud would save money but to most people it has turned into a financial abyss. The pay-as-you-go model is also a good idea until you recognize how quickly the micro-transactions can add up.

Indicative example, one SaaS business with annual revenue in the mid-market range, that we audited was paying a quarter of a million dollars every month to cloud providers, until they found out that they were being charged 40 percent to run non-utilized resources.

  • Concern of Gartner: More than 60 percent of companies overspend on cloud because of the lack of cost control.
  • Tech insider: Cloud invoices are minibars in hotels, they are very convenient, but every pack of snacks costs a fortune at the end of the bill. CloudZero — FinOps Specialist

The irony? The cost of running own servers, including maintenance, is now found to be cheaper in the long run to some companies.

Security & Compliance: When Control Matters More Than Convenience

Public clouds are safe- until they are not. One of the big healthcare organizations we have interviewed (who do not wish to be named) transferred patient records back on-prem when they nearly suffered a data breach because they had an S3 bucket misconfigured.

  • Regulatory pressure: EU’s Digital Markets Act and China’s data sovereignty laws are forcing firms to rethink where data lives.
  • The Goldman Sachs model: The bank uses on-prem trading algorithms to maintain sub-millisecond latency and compliance with the regulatory requirements.
  • Professional opinion: When data holders consider their data as their crown jewels, would they prefer to have it in a vault of other parties? –CISO, tech fortune 500 company

In such industries as finance and healthcare, the shared responsibility model offered by the cloud is not only risky, but it is legally risky.

Performance: The Cloud’s Achilles’ Heel

Latency could not give a hoot about your uptime SLA. This lesson was taught to our gaming studio client when cloud-based rendering added 200ms, which ruined the player experiences. They went back to on-prem GPU clusters and the frames rates stabilized immediately.

  • The evidence does not lie: IDC discovered that 45 percent of IT leaders identify unpredictable performance as one of their cloud strengths.
  • Hypothetically, edge spin: Hybrid systems (such as the on-prem AI that Walmart uses to track its inventory) combine the scalability of the cloud with a high computing capability.

The lesson? Not every work load has to be in another persons data center.

The Hybrid Future: Best of Both Worlds?

Companies that are the most intelligent are not leaving the cloud: the trend is getting picky instead. NASA, as another example, has reserved the Mars rover data on proprietary servers but has used the AWS to store portals accessible to the general population.

  • Tooling shift: VMware gives companies such as BMW the ability to train ML models in-house and use the cloud to do inference using a product called Private AI.
  • Smart forecast: The workload-aware infrastructure is the next generation The next 10 years are the workload-aware infrastructure, not dogma. Red Hat, CTO

It is not about denying the cloud. It is the question of intentional use.

Conclusion: The Cloud’s Reality Check

The cloud transformed IT – it is not a religion. The cheapest, the most regulatory, and the least performing will be the questions: Where does this work load actually belong?

Thinking experiment: Why do you keep it all in your cloud, even though you would stop using it next day, should your cloud bill go up twice? The answer to this may shock you.

Fire starter: Have your company repatriated workloads? Commenting is open like war stories.

Leave a Reply

Your email address will not be published. Required fields are marked *