Cloud Repatriation: When Is it Time to Bring Workloads Back On-Prem?
It feels like we’ve been talking “cloud-first” or “cloud-only” when it comes to IT transformation and new procurement strategies for years now. But in 2019 we already saw some signs of what analysts are now calling cloud repatriation. At the time we asked, are enterprises moving back to on-premise data centers?
The answer isn’t that simple, but there is certainly a time and place for cloud repatriation. Here’s why it’s trendy to move some workloads back on premise and how to decide whether its time for you to follow suit.
Is cloud repatriation a real trend?
Yes, enterprises really are moving significant amounts of their workloads back on premise. One IDC survey from 2018 discovered that 80% of respondents moved cloud workloads on premise or to a private cloud solution within the past year. It also found they planned to migrate an additional 50% of their current public cloud workloads either on premise or to a private cloud within the next two years.
Why is this happening? After all, the advantages of cloud computing have been well-established, including increased agility and plug-and-play PaaS components for everything from security to analytics to automation.
The reality of cloud costs have kicked in for some deployments. Reserved instances of hyperscale cloud resources can add up for large scale enterprise computing. This is especially true for so-called “lift and shift” type migrations in which legacy apps are moved wholesale into public cloud with little regard for refactoring or rearchitecting with cloud advantages in mind.
Survey respondents also reported performance and downtime as issues, as high profile outages become seemingly more common. Of course, an on premise solution doesn’t negate downtime — but it does introduce a sense of control over the infrastructure. At least it will be your fault if your data center goes down.
Along with that control factor comes security. Some enterprises (and more importantly their shareholders) may feel that public cloud doesn’t offer the kind of security and compliance controls that an on premise or private cloud solution might.
What solutions are emerging?
It would appear that the hyperscale providers see this trend continuing as they extend their hybrid cloud offerings into on premise data centers. One prominent example is Azure Stack. AWS has a similar offering. The gist is a hardware stack that resides in your datacenter, or using a combination of pre-approved hardware and software provided by the cloud service provider. This technology runs the same cloud platforms as their public cloud services, so you can seamlessly extend your on-premise infrastructure to the cloud.
While this does provide some private cloud benefits, it might not satisfy all security and compliance measures. It also is very limited in terms of the technology you can leverage.
Good old colocation offers a best of both worlds approach in which you do not need to manage the underlying support infrastructure for your servers, storage, and network equipment, but you can still combine the technologies you need for your private cloud solution.
By working with a knowledgeable MSP who can offer colocation, data center modernization, and hybrid cloud services across a wide array of platforms and providers, your IT team can minimize the headaches involved in administration while still gaining the benefits of a true hybrid solution.
When should you consider cloud repatriation?
There are a few situations in which you may want to move some or even all of your infrastructure back to an on premise or private cloud deployment. The first, as mentioned above, is if your public cloud expenses have crept beyond what an on premise solution would cost. Dropbox and New Belgium Brewing are two high profile examples of organizations who have repatriated their primary applications.
Hyperconverged infrastructure is one development that has nudged enterprises in this direction. It greatly simplifies the deployment and management of data centers as storage, network, and compute are all bundled in one software-defined package, so they can be administered from a single platform and installed in a single rack.
Steady workloads can be less expensive to run on premise if you have the skills and time to set up efficient automation and software defined technologies. Colocated, hyperconverged infrastructure may be easier to control for performance and troubleshooting as well — again, if you have the IT staff bandwidth to do so.
Workloads that require low latencies or that transfer large quantities of data over the network are also prime candidates for repatriation, as network transit costs can make up a significant portion of a cloud bill.
Finally, security and compliance might be simplified with a private cloud, on premise, or colocated environment. The audit trail is easier reference and you have more granular security control.
When to stay in the cloud
On the flip side, the cloud offers baseline performance and ease of use when it comes to security and PaaS components. It remains relatively simple to provision new virtual resources, add the networking, storage, analytics, and automation components that you need, and control it all via a single portal.
It is certainly much, much faster to spin up new servers using the cloud than to order, install, test, and deploy them in a data center. This means that for unstable, changing, or relatively new workloads, the cloud remains the best place to deploy.
If you plan to design and build your own data center, costs can be very prohibitive, especially for the kind of redundancy enterprise applications require. Even with colocation, you can feel stuck in a way that does not exist in the cloud. After all, you invested thousands of dollars in IT equipment, so you can’t switch things up on the fly.
Various dependencies on cloud technologies may not be so simple to refactor for your on premise infrastructure, either. Essential pieces, like network configuration, may require serious work from your IT team to get working again.
Whichever decision you make for your primary applications, it is likely that a hybrid deployment will remain the best solution moving forward. The key is ongoing cost modeling and analysis so you can remain informed on whether public cloud or private deployments are best for individual workloads and data storage scenarios.