RISKS
Due to its highly distributed and diverse nature, multi-cloud adoption is not without its share of risks. For example, with the onset of the Covid-19 crisis, issues around automation, security, and visibility across multiple clouds quickly came to the fore, said Quinn. “As organizations expand to multiple clouds, their network is spread across additional providers and it results in blind spots that make it difficult to manage devices. This presents a huge risk to organizations, especially since the number of IoT devices has grown exponentially due to remote work.”
Having a multi-cloud strategy is similar to “having multiple data centers, each … architected differently in terms of deployed tools, compute platform, storage platform, and networking, which increases administrative overhead,” said Rulf. “You will need to extend your management and monitoring tools to support the different APIs, native tooling, and each cloud platform.”
And “with multi-cloud strategies in place, organizations might experience complications like framework misalignment and differing security policies,” said Tareen. Vanover also warned that a “side effect of today’s multi-location, multi-data center, multi-cloud strategy is hyper-sprawl, meaning that there is federation of data everywhere, across multiple clouds, databases, and devices.” Moreover, there are risks of “multi-cloud inefficiencies in some organizations, especially around purchase power, cloud overspend, and a potential increased security burden.”
In addition, pointed out Chander Damodaran, CTO of cloud engineering at Brillio, the “complex architecture required to address multi-cloud resilience is potentially fraught with higher costs, complicated delivery, and potentially longer time to market.” Another issue is the technical skills and resource availability needed to manage multi-cloud environments.
Observability is another requirement for multi-cloud sites, as enterprises will need to ensure performance and security from a number of vendors. “Without an observability platform in place, organizations attempting a multi-cloud approach are flying blind and can introduce a host of risks,” said Tareen. “It’s like turning off your radio before trying to land a jet at JFK. Make investments upfront to ensure you remain in control and have a clear, safe runway for your cloud applications.”
A multi-cloud deployment “will, most likely, require more operational effort to monitor and maintain,” said Burns. “Just as single-cloud deployments are handcuffed to changes on the platform they run on, multi-cloud deployments are subject to any changes to two or more different platforms. As a result, multi-cloud deployments face potentially exponentially increased workloads. Resource dependency between detached components can also become complicated, especially in data engagements where data sources are often being utilized across clouds. These connections are usually accomplished by using platform-specific connectors, which inherently tie down a multi-cloud deployment as well.”
TOOLS AND TECHNOLOGIES
When it comes to managing data across multiple clouds, “it is essential to look for tools that will give the user access to all platforms that host data—whether that’s private cloud, public cloud, or on-prem—providing one centralized API to manage in-band and out-of-band operations,” said Speciale. “The platform should be cloud-provider-independent, meaning users decide on the optimal cloud storage option for their data. The platform should act as an abstraction layer between data and the proprietary APIs used to place that data in any given cloud storage service. It should also be able to translate data into the native format of any given cloud storage service so that the data remains open and accessible to any system and not just the system that placed the data. This will prevent time-consuming and costly migrations and pave the way for a more flexible and efficient workflow to keep data more available and actionable.”
The rise of containers has also paved the way for multi-cloud adoption, industry observers said. The correlation between multi-cloud and containers is strong, since the ultimate goal of deploying software in containers is portability across environments such as on-premise (bare-metal servers) and cloud environments, said Speciale. “In the past, when we wrote monolithic applications for specific target environments—Windows, Unix, Linux—there was really no hope of portability. Application developers would need to work hard to port the application to a new environment. Now that we have users, applications, and data being deployed using data center resources and the cloud, a central goal is to make applications intrinsically portable. If we really want to reap the benefits of multi-cloud, we absolutely will need the software to run anywhere, and that, in turn, is one of the key benefits of containers.”
Increasingly, enterprises “are betting on Kubernetes container technologies to drive their multi-cloud strategies and are deploying applications to easily orchestrate, manage, and scale while also retaining the flexibility to port applications across different clouds,” said Mathur.
This convergence is not without its issues, however. “Orchestration is necessary for any sort of cloud deployment at scale, whether multi-cloud or not,” said Rulf. “That said, I have not seen container platforms easily deployed in a multi-cloud strategy. Most cloud platforms have unique feature or functionality built into their container platform that prohibits doing container deployments that span across cloud providers.”
To ensure greater transparency and control, Quinn advised putting a “cloud manager” in place to provide visibility across multiple cloud environments. “A good cloud manager combines data governance, data catalogs, and enterprise metadata management capabilities, providing customers visibility into their data landscape.” In addition, Rulf urged administrators to adopt robust monitoring and billing analysis tools to increase transparency of multi-cloud arrangements.
When enterprises embrace a multi-cloud strategy, “it becomes critical that data and applications are always available across all cloud types, including private, public, and hybrid clouds, to meet innovation and competitive demands,” Vanover said. That means enterprises must ensure availability is at the forefront of their digital transformation strategy, so that when applications and workloads are being moved across different infrastructures, there’s a backup and disaster recovery plan in place so that downtime does not become an issue.
THE GREAT DECOUPLING
As enterprises move to multi-cloud environments, they need to understand which applications—or parts of applications—may be best-suited to deployments across multiple cloud vendors. Decoupling applications from underlying platforms may ensure the flexibility required. “When viewed through the lens of application modernization, new, modernized applications are really just collections of components deployed on top of a platform orchestration fabric,” said Burns. “These modern application components include containers, serverless functions, storage, relational databases, and non-relational databases.” Understanding which of an application’s components are spread across a hybrid environment is key to understanding cloud scale, Burns said. “It also points to why application decoupling will unlock the most potential from your cloud deployment.”
This changes the way applications operate as well, Burns continued. “You likely can’t span traditional monolithic applications across multiple service fabrics and still operate as expected,” he said. “The emergence of multi-cloud deployments has expanded the definition of cloud scale to include separately deployed components of an application. Applications deployed within your own data center, for example, can now leverage public cloud databases.”
Ultimately, flexibility and choice will encourage more competition, opening the door for innovative providers and solutions that are capable of running on the three major cloud platforms and incorporate private cloud and on-premise technologies as needed, said Mathur. More competition “will also encourage the dominant public cloud providers to double down on open interfaces, upgrades, and innovations to ensure their offerings integrate seamlessly with more platforms. The move toward multi-cloud adoption with an eye toward portability and compatibility across cloud platforms not only allows companies to select from best-of-breed solutions for improved business outcomes but also gives them more control over their cloud environments.”