The Goldilocks Principle is the premise that we tend to end up with “just the right amount” of something over time. The Goldilocks Principle shows up in many places in our lives and business in areas as vast and varied as astrobiology, economics, and machine learning. When it comes to the cloud, I can place my twist on how the principle is applied.
In my twist, the high-end offering is "all in with the public cloud," not an option for most companies. In recent years we have seen companies repatriate back from the "all in" scenario or, at the very least, place bets across multiple cloud providers.
The low-end offering is to retain everything on-premises. Sometimes, there is no choice because of regulatory compliance, security requirements, or performance concerns.
The hybrid cloud is my middle, “just the right amount” offering. Every company has unique requirements but taking advantage of everything the cloud offers while retaining a partial on-premises footprint for proprietary reasons is the direction of the majority.
If “just the right amount” is the answer – why is it taking so long?
The public cloud is eighteen years old, and, depending on whom you talk to, only 10-25% of applications are in the cloud. Many of you will recall BoD and C-Suite mandates such as "we will be all cloud by 2017", "no more spending on anything on-prem," and (my favorite) "we're going to save 50% of our cost in the cloud".
Private clouds came on the scene in 2008, largely undeveloped and definitely in the “do-it-yourself” product class. I'm sure many of you struggled with OpenStack. I did not find anyone who could get it to work apart from the biggest telcos and retailers.
When vendors introduced hybrid clouds around ten years ago, the interoperability between a private and public cloud and the ability to move workloads back and forth was a science project. Very few companies had the resources to pull it off.
Then about five years ago, we saw the change. At that time, the advantages of the cloud (OPEX, speed, agility, flexibility) were well known, as were the disadvantages (Cost creep, latency, data concerns, teams for each cloud).
While Google Cloud and Azure committed early to the hybrid cloud with Anthos and Stack, the last big hybrid shoe to drop was AWS, which now offers 21 Hybrid Cloud services, and we were off to the races.
Although not as mature as the 18-year-old public cloud, there are many hybrid third and fourth-generation offerings from all the major players, such as Google (GCP), Azure, Oracle, IBM, HPE, Lenovo, and VMware.
Now is the age of hybrid multi-cloud
First, let us be clear on terminology. A hybrid cloud is a combination of public and private clouds.
The most important feature of a hybrid cloud system and the critical inhibitor of growth in the past is the orchestration, management, and application portability across private and public clouds. A hybrid cloud is a single, unified, and flexible distributed computing environment which can run and scale traditional or cloud-native workloads on the most appropriate computing model.
A multi-cloud environment includes multiple public cloud services from different cloud providers. Apart from spreading wealth, the primary reason for adopting a multi-cloud strategy is that not all providers are created equal. A company might host its web front-end application on AWS, Exchange servers on Microsoft Azure, and machine learning applications on GCP.
A hybrid multi-cloud is a hybrid cloud with public cloud services from more than one provider.
The “best of both worlds”
A hybrid cloud offers cloud services when it works best while keeping certain operations on-premises in a private cloud. Combining the control of an on-premises data center or private cloud with the scalability of the public cloud means that each workload runs in its optimal environment for performance and total costs. Workloads that require a high level of control, customization, or security can run in an on-premises private cloud. In contrast, other workloads can take advantage of the scale and flexibility of the public cloud.
The decision of where to place a workload depends on many factors, such as economics, security, resilience, and Environmental, Social, and Governance (ESG). The mix of elements results in a “workload personality” that best fits a particular infrastructure. Over time the factors and hence the personality change resulting in continuous shifts in optimal deployment.
A hybrid cloud provides the portability to move workloads from on-premises to public clouds and back again as requirements change. A changing "workload personality" is well-serviced by a hybrid cloud.
In summary, the hybrid cloud benefits from the control and visibility over the private cloud, with the flexibility, scalability, and reduced capital cost of the public cloud.
Critical success factors for a successful hybrid cloud
At this point, most articles on hybrid cloud end abruptly. But, let us continue to discuss some critical success factors.
Managing data lifecycles across a hybrid cloud environment can be tricky. By data lifecycles, I mean collecting, reporting, enriching, and serving data to analytics and machine learning (ML). Administrators must use tools and APIs to ensure analytics can be run across private and public clouds without disruption, requiring the separation of data management from infrastructure strategy.
A hybrid cloud must have unified metadata, data access, governance, and lineage across private and public clouds through a user interface, regardless of data history - sourced, migrated, or replicated. If data access policies and origin are inconsistent across private and public clouds, gaps will exist in audit logs, creating an information security and compliance nightmare.
The third and final factor that makes a hybrid cloud possible is having an open architecture, open-source software, open APIs, and an open storage format. Open source enables interoperability because the same services run in the private and public clouds. It also improves portability and removes the lock-in dependencies to any one cloud provider's business model and priorities.
Managing a hybrid cloud
Hybrid cloud management is complex. It must be a holistic solution containing data management capabilities to align the three success factors described above. Look for a management platform that supports the entire data lifecycle for analytics and ML across clouds, provides data security and governance, and is based on open-source software. The management platform must keep data security and governance policies consistent everywhere between private and public clouds.
To make a hybrid cloud a reality, data platform administrators need one convenient user interface to administer, manage, and provision users, environments, and services across all clusters and analytic workspaces. Administrators must know the "workload personalities" and the relationship to on-premise and public cloud resources.
Accelerated product development cycles require a streamlined, step-by-step process to migrate workloads and burst to the cloud. It also requires transparency into the health of clusters and workloads.
What comprises a “just right” solution? The Goldilocks Principle predicts a trend toward the "just right” balance between two extremes, which is indeed happening. Every enterprise needs a strategy that encompasses the hybrid cloud reality.
A hybrid cloud allows interoperability across on-premises and cloud. The same level of access and flexibility must be available for data. Workloads or datasets must be portable in a hybrid cloud to respond to evolving needs. Where applications and data reside today might not be the best place for them tomorrow. Every hybrid cloud strategy must have a companion hybrid data strategy.
As aptly stated by Sudhir Menon, Chief Product Officer at Cloudera. “We’ve heard loud and clear from our customers and the industry that data management tools and platforms that fail to support hybrid and multi-cloud capabilities will be set for decommissioning.” Or stated differently, "ignore the Goldilocks Principle at your own risk."