The hybrid cloud balance: Knowing when to shift between public and private

In the last few years, industry analysts have been discussing the phenomenon of companies considering taking their workloads off the public cloud. In fact, a recent argument that market capitalizations of scale public software companies is weighed down by cloud costs, and by hundreds of billions of dollars, caught the interest of several enterprise leaders. It is easy to misinterpret this as a prediction of an imminent exodus from the public cloud, which I doubt will be the eventual turn of events. Data shows that only a modest number of companies — a 2019 survey by Gartner put that number at 4% — have actually repatriated (or need to repatriate) their public cloud workloads to a private cloud solution.

My own view is that the public cloud is indispensable to digital transformation. It remains one of the biggest opportunity areas for organizations and is practically the only proven way to scale a business quickly and reliably. And yet the approach to cloud that promises most value for enterprises is hybrid — both public and private — leveraged for the right reasons and at the opportune moment in a business’s lifecycle.

Often, disillusionment with the public cloud, and consequently advocacy for repatriation, stems from a misestimation of cost savings from cloud migration in the first place. Contrary to popular notion, cloud’s promise of affordable compute and storage is not the strongest driver of operational efficiency for enterprises — especially not as growth often slows with scale, unit costs build up, and diminishing near term efficiency starts to sound the alarm bells. Not all workloads are suited to the public cloud either. Very often, the problem is not the cloud itself but poor workload planning and management along with misplaced goals. For example, enterprises that think they can simply lift and shift their on-premise workload to the public cloud are often disappointed to find that the path to sustained value isn’t paved exactly that way.

However, there may come a time in an enterprise’s hybrid cloud journey when, depending on business motivations, it makes sense to shift the business to lean on public or private cloud. This is when the enterprise is significantly cloud-mature and may be looking at optimizing its workload for a number of reasons:

To lower cost: From an efficiency point of view, it takes the complete automation of operations built atop the cloud infrastructure layers of compute and storage to take advantage of cloud. Relying simply on cutbacks in upfront capital expenditure along with flexible running costs of infrastructure, commensurate with growth in operations, places an unsustainable burden on the business as users and operations scale grow. This is a tricky situation, since the business cannot scale, first off, unless it’s on the public cloud, but once it achieves a certain scale it can no longer afford the running expenses if its IT and business operations are not automated. Another important source of value on the cloud is the ecosystems — the relevant platforms and marketplaces that promise enterprises disproportionate efficiencies inaccessible to competitors who may not be harnessing the cloud similarly.

If, however, scaling business is the predominant motivation for a business to embrace the public cloud, on achieving critical scale, the enterprise may consider repatriation to a private cloud, in part or full, to take control of costs. Dropbox offers us an interesting example of a company making the switch on the path forward. Dropbox exited the public cloud (for the most part) in 2016 to put all its data in three owned colocation data centers across the United States. In the first year, the company shaved U.S. $92.5 million off its direct billings; even after accounting for the costs of building the new facilities, the savings were nearly U.S. $75 million. But in order to grow to 500 petabytes Dropbox relied first on the public cloud.

The public cloud also brings with it the opportunity to harness a whole data economy beyond one’s own enterprise data. Explorations to bring the best of public cloud services to the private cloud are now ongoing at scale. This will enable businesses to gain the cloud experience and value benefits while retaining control of their data to meet data governance and residency regulations.

Beating latency and improving availability: Public cloud service unavailabilities are rare but can create large-scale disruption when they do occur. At that time, client enterprises can do little but wait. But with a private cloud or local data center, an enterprise has more control over availability and downtime. For example, health care providers who store patient records on the public cloud, and patient monitoring device data, could be retrieved on demand from the private cloud. Latency can also be an issue, especially if the user base is massive and geographically distributed. It is precisely to overcome this that Netflix runs on a hybrid cloud model, hosting its content and user database on a public cloud, while streaming content locally to users through its private Content Delivery Network.

Security comfort: Technology-wise, the public cloud is highly automated for security, which means less human intervention and fewer errors. Cloud security may offer specialized options otherwise out of reach, because of costs, for many enterprises. Often, cloud-based security services are pre-configured, and if the enterprise prefers that the system be set up differently, there may not be many options. This, on occasion, drives enterprises towards the private cloud, especially out of consideration of the regulatory environment, reporting requirements, or data sensitivity — think banks and healthcare organizations.

Lack of skills: It takes deep skills in the areas of provisioning, cloud architecture, and performance reporting, to name a few, to manage workloads efficiently when running a private cloud. Those relying on the public cloud get a lot of help on this front due to the provider’s automation. For example, Google’s Anthos is a managed application platform that extends Google Cloud services and engineering practices to any environment, even outside of Google Cloud Platform, so enterprises can modernize apps faster and establish operational consistency across them. So for businesses that face a skills shortage, the private cloud or local environment may not be viable options.

Clearly, repatriation is not a “mass” option, nor the default move at a defined stage of an enterprise’s cloud maturity. At any point, the right workload should be housed, for the right reasons, at the right time, on the right cloud — public or private. The value of AI-led automated operations, data exchanges, and software-as-service made possible on the public cloud is not to be dismissed easily. However, every organization will do well to factor repatriation into its long-term contingency planning as a possible response to future changes within its own business and also to the public cloud.

Ravi Kumar is President at Infosys.