Traditional hybrid cloud architecture

Initially, hybrid cloud architecture focused on the mechanics of transforming portions of a company's on-premises data center into private cloud infrastructure, and then connecting that infrastructure to public cloud environments hosted off-premises by a public cloud provider (e.g. AWS, Google Cloud Services, IBM Cloud, Microsoft Azure). This was accomplished using a prepackaged hybrid cloud solution such as Red Hat OpenStack (link resides outside ibm.com) or by using sophisticated enterprise middleware to integrate cloud resources across the environments, and unified management tools for monitoring, allocating and managing those resources from a central console or 'single pane of glass.'

The result was unified IT infrastructure well-suited to several use cases:

• Security and regulatory compliance: Reserve behind-the firewall private cloud resources for sensitive data and highly regulated workloads and use more economical public cloud resources for less-sensitive workloads and data
  
  
• Scalability and resilience: U​​se public cloud compute and cloud storage resources to scale up quickly, automatically and inexpensively in response to unplanned spikes in traffic without impacting private cloud workloads (this is called 'cloudbursting)
  
  
• Rapid adoption of new technology: Adopt or switch to the latest software-as-a-service (SaaS) solution, and even integrate those solutions into existing applications, without provisioning new on-premises infrastructure
  
  
• Enhancing legacy applications: Use public cloud services to improve the user experience of existing apps or to extend them to new devices
  
  
• VMware migration: 'Lift and shift' existing on-premises workloads to virtualized public cloud infrastructure, to reduce the on-premises data center footprint and scales as needed without additional capital equipment investment
  
  
• Resource optimization and cost savings: Run workloads with predictable capacity on private cloud and migrate more variable workloads to public cloud; use public cloud infrastructure to quickly 'spin up' development and test resources as needed

Modern hybrid cloud architecture

Today, hybrid cloud architecture is focused less on physical connectivity, and more on supporting the portability of workloads across all cloud environments, and on automating the deployment of those workloads to the best cloud environment for a given business purpose. Several trends are driving this shift.

As part of the next critical step in their digital transformations, organizations are building new applications and modernizing legacy applications to leverage cloud native technologies—technologies that enable consistent and reliable development, deployment, management and performance across cloud environments and across cloud vendors.

Specifically, they're building or transforming applications to use microservices architecture, which breaks applications into smaller, loosely coupled, reusable components focused on specific business functions. And they're deploying these applications in containers—lightweight executable units that contain only the application code and the virtualized operating system dependencies required to run it.

At a higher level, public and private cloud are no longer physical 'locations' to connect. For example, many cloud vendors now offer public cloud services that run in their customers on-premises data centers; private clouds, once run exclusively on-premises, are now often hosted in off-premises data centers, on virtual private networks (VPNs) or virtual private clouds (VPCs), or on dedicated infrastructure rented from third party providers (who are sometimes public cloud providers).

What’s more, infrastructure virtualization–also called infrastructure as code–lets developers create these environments on demand using any compute resources or cloud resources located behind or beyond the firewall. This takes on added importance with the advent of edge computing, which offers opportunities to improve global application performance by moving workloads and data closer to where the actual computing gets done. 

As a result of these and other factors, modern hybrid cloud infrastructure is starting to coalesce around a unified hybrid multicloud platform that includes:

• Support for cloud-native application development and deployment across all cloud types (public and private) and cloud providers
  
  
• A single operating system across all environments
  
  
• A container orchestration platform–typically Kubernetes–that automates the deployment of applications across cloud environments

Cloud-native development lets developers transform monolithic applications into units of business-focused functionality that can be run anywhere and reused within a variety of applications. A standard operating system lets developers build any hardware dependency into any container. And Kubernetes orchestration and automation gives developers granular, set-it-and-forget-it control over container configuration and deployment–including security, load balancing, scalability and more–across multiple cloud environments.