Losing Sleep Over Public Cloud Bills? Time to Consider Repatriation

When image editing and sharing app Picsart got a bill from Google Cloud Platform (GCP) three times higher than normal, they were shocked. Worse: the line item responsible for the pricing surge was described only as “data transfer.”

But the final straw was when GCP refused to answer Picsart’s calls unless they upgraded to an enterprise account. At the time, the app company was spending hundreds of thousands of dollars a month with the hyperscaler.

It’s an infuriating scenario, but one that’s becoming common for growing companies hosted entirely on the public cloud. In fact, 93 percent of senior-level IT professionals are now intentionally combining public and private cloud in their hosting setup.

Their top priority for the next three years? Building workloads in the private cloud.

If any of this resonates, you (like Picsart) are likely considering cloud repatriation as a way to bring greater predictability to your hosting costs and future-proof your most important workloads. In this piece, we’ll explain how bringing workloads off the public cloud with help from a managed services provider can help you do exactly that.

Better Performance, Predictable Costs

One thing Picsart liked about GCP was the flexibility. During 2020’s pandemic quarantine, the app saw a surge of usage that its hyperscale host handled elegantly. But as they introduced more AI and machine learning features, the cost of hosting in the public cloud kept climbing.

On-prem hosting didn’t seem like a viable alternative, however: GPUs are expensive and hard to get. They have a short shelf life. The Picsart team knew it wasn’t viable for them to invest so much time and money in managing an AI-capable data center.

When they started considering managed services partners to help them migrate from the public cloud to a hybrid setup, they looked for the following capabilities:

• Scalable architecture: The ability to spin up (and down) GPU nodes and storage quickly without compromising performance.
• Resilient infrastructure: Excellent power, cooling, and failover support so performance didn’t drop when demand spiked.
• Security baked in: This includes everything from multi-factor authentication to disaster recovery – key if you operate in a highly regulated industry or are handling sensitive data.

And then there were the cost considerations.

As industry analyst David Linthicum points out in a recent InfoWorld article, “public cloud providers are not incentivized to dramatically lower costs,” despite the falling costs of the hardware they use to provide their cloud services.

Why? Lowering costs could hurt shareholder returns, for one thing.

For another, hyperscalers’ business models are highly complex and they offer a variety of ancillary services that can benefit cloud customers.

One public cloud alternative Linthicum highlights is managed services providers. But MSPs aren’t just a more cost-effective option – they also deliver essential support the hyperscalers generally don’t.

Guidance, Support, Advice – When You Actually Need It

As Picsart learned the hard way, GCP and other hyperscalers require a huge financial commitment before they’ll offer expert guidance and human support.

MSPs are the opposite.

At Summit, for example, we offer live human support 24/7/365. After a while, you might even recognize the voice of the person who picks up the phone.

And this isn’t just something we offer when things go wrong.

From day one, we’ll work with you to develop a migration plan that makes sense for your organization and your growth goals.

This is huge for a lot of companies looking to repatriate workloads to the private cloud. As many as 30 percent of IT leaders cite a lack of internal skills as a barrier to private cloud adoption. In other words: they’re not making the move because they don’t have the know-how to make it work.

What's more, 80 percent of in-house teams rely on professional service providers (like us) to handle their cloud-related needs.

That’s only natural given the way the industry has evolved. Most people entering IT in the last decade or so have only ever worked in a cloud-first world. Evolving to a hybrid posture will naturally require support from experts who have been working with private cloud, colocation, and other hosting setups all along.

One word of caution, however: as you choose the partner to guide your repatriation, make sure they’ve got a track record of success.

Warning: Don’t Be the Crash Test Dummy

This is especially true if you’re looking to repatriate AI workloads. Many MSPs say they can provide an AI-ready hosting environment—and it may look like that on paper.

But looks can be deceiving. 

The truth is that AI is still relatively new—new enough that many providers haven't actually put rubber to the road when it comes to their AI promises. The last thing you want is to be the crash test dummy strapped in on their first run.

That's why it's important to make sure you're choosing a provider that's already managed AI workloads. More specifically, you want a partner that has already deployed every major GPU type, understands how to navigate the pitfalls that come with AI scaling, and can handle the power, cooling, and storage demands these systems introduce.

That's where Summit stands out. Our team has enough experience to support your AI ambitions at any stage. We'll help you figure out your steady state workloads, choose the right home for your data (e.g., hybrid cloud), and help you optimize deployment.

As you’re researching partners, ask about past successes with work similar to yours.

For example, consider the AI recruiting platform Summit recently helped repatriate from AWS, where their costs were skyrocketing as they reached a steady state.

With the hybrid solution we devised for them, they enjoyed…

• 30% cost reduction from their AWS spend.
• 50% more processing power compared to their AWS setup.
• Ample room to grow without increasing spend.
• Support from human engineers rather than AI chatbots.
• Noticeable performance improvement, thanks to local storage (vs. shared network storage from AWS).

Future-Proof Your AI Hosting with a Managed Services Provider

The public cloud is a great place to experiment with AI workloads. It’s quick, low-commitment, and scalable as you figure out what works.

But once you enter a steady state, the public cloud gets expensive and difficult to manage. That’s only going to get more true as you scale up. If you’re looking for ways to optimize your cloud investment, ensure peak workload performance, or otherwise future-proof your most critical workloads, it’s time to consider what an MSP can offer.

We’d love to hear what's keeping you up at night right now. If it’s related to your public cloud usage, there's a good chance we can help you sleep better.