Skip to main content

In the last few months there has much written about cloud repatriation – migrating workloads back from public cloud to an on-premise environment. But after the time and money spent migrating to cloud, why on earth would you want the upheaval of moving back? 

In this post, we consider what is driving repatriation and whether it is the future, a fad or just fake news. 

What’s driving public cloud users to move back on-premise? 

Supporters of repatriation claim that they’re able to save significant cost on-premise versus operating workloads on cloud. Dropbox, often heralded as the pioneer of repatriation, claim to have saved $75 million over 2 years by doing so.  

Given cost reduction is one of the main drivers for shifting to public cloud, you might ask yourself ‘how were they able to save so much money by NOT moving to public cloud and why aren’t we doing the same thing?’. 

Dropbox and other successful repatriations all share similar characteristics and follow a similar path. The organisation builds and scales their application in public cloud before reaching an inflection point where they’re able to command greater economies of scale for buying and operating their own hardware which offsets the cost of the margin charged by the public cloud provider. Those who’ve succeeded when it comes to repatriation have had workloads that are both massive and stable. 

Dropbox is a cloud storage service, users upload their data to the platform and access it when they want, it’s predictable and always available. For these workloads, cloud cost optimisation techniques, like spot instances, reserved instances and autoscaling can’t be effectively deployed. However, for workloads that can make use of these cost optimisation techniques, even organisations like Dropbox still use public cloud, effectively adopting a hybrid model. Don’t necessarily believe all the hype –cloud repatriation doesn’t necessarily mean everything is moved back on-premise. 

So, for some very specific workloads, running an on-premise environment seems to be cost-effective when compared to the cloud. But is it really that simple? 

Potential problems on the horizon 

Imagine you’re back on-premise. Capacity demand suddenly grows and you need more compute in a hurry. At which point, you’re back to the pre-cloud conundrum. You need to invest in on-premise resources to meet fluctuations in demand which then sit redundant if demand drops. This level of investment perhaps wasn’t considered when you decided to move back from the cloud and may mean the move isn’t as cost-effective as you thought. 

You need to think about the CapEx you’ll need to ensure your on-premise infrastructure can scale as your business grows to meet future demands. But that’s not all. Let’s not forget that the main reason companies move to the cloud is to outsource the problem (and distraction) of data centre construction, maintenance, security and support in order to focus on their core business. 

Repatriation isn’t as simple as just flipping applications back to your on-premise facility. You’ll need to migrate to that location, re-train staff and implement resiliency plans. No small undertaking. 

So, should you consider cloud repatriation? If you’re SaaS provider who operates a giant, stable and predictable application which uses no cloud-native services, it may well work out. Especially if you don’t foresee significant rises in demand any time soon and you have the infrastructure skills in-house to make the move. Do the math. If the numbers seem to say repatriation may work out, then you may be left scratching your head and doubting the cloud. 

Regardless of the profile of your workloads, before you decide to make any moves, we advise you to review whether you’re using cloud effectively first. Only then can you make a fair comparison. 

It’s not you, it’s me 

Failure to hit the bullseye isn’t the fault of the target and simply moving the target won’t improve your aim. If you’re not seeing value or efficiencies from using cloud compared to running on-premise, could it be that cloud isn’t the problem? Maybe you need to look at whether you’re using cloud to its full potential.  

High cloud bills are often self-inflicted by enterprises that don’t refactor applications and data to optimise their cost-efficiencies for cloud. If you lift and shift in bulk, it’s not just the applications you shift to the cloud – you’ll also transfer any inherent inefficiencies. Cloud bills will be higher than expected because lifted-and-shifted applications can’t take advantage of cloud native capabilities such as auto-scaling and storage management that allow workloads to function efficiently. 

Cost isn’t the only consideration when it comes to cloud and not the only benefit it can deliver when used properly. Leveraging cloud can deliver agility, scalability, security and resiliency, not to mention improvements in the pace of delivery and service innovation. Harnessing these benefits requires changes to governance and your approach to risk management which is arguably more difficult than moving applications to the cloud.  

Is it time to take control? 

At Cloudscaler we’ve seen cost optimisation techniques drive a 40% year on year drop in cloud costs. By using a control framework, landing zone and effective cloud operating model, enterprises can optimise and accelerate their time to cloud value. If you are convinced public cloud isn’t right for you then you might want to consider some of the on-premise private cloud solutions available today which might enable you to reduce the ‘distraction factor’. 

So, before considering a move back to on-premise, we advise businesses to ensure they are really comparing like-for-like costs – gather and analyse the relevant data to identify whether you have an issue and what the root cause might be. Armed with that data you can make an informed decision that is right for your workloads and your business rather than believing the hype. Changing car may not be the answer when it comes to transformation success, you may just need to tinker with the engine.