Multicloud was a natural part of the picture from the first days of cloud computing. Reducing on-premises data center resources meant cost savings,  and spreading cloud applications across several cloud providers mitigated risks.

Integration between different cloud platforms was on many IT roadmaps. It still is. In fact, Cisco’s recent 2023 Global Networking Trends Report indicates that  92% of those surveyed are using more than one public cloud.

[ Related: What is cloud computing? ]

Using multiple clouds from different vendors reduces lock-in, provides leverage in price negotiations and allows IT heads to assign special applications to best-of-breed services providers. Those benefits will always find takers.

But we're still in the early phase of multicloud implementations. Obstacles to rich integration of multicloud applications still stand today,  almost two decades after the advent of cloud computing. And that's no surprising:  multicloud adds more ingredients to an already-demanding integration stew, and further innovationa are always forthcoming. Planning, designing and architecting tomorrow’s multicloud calls for far-sightedness.

Of course, enterprise application integration has always been challenging. While loosely coupled, cloud-centric development approaches bring greater flexibility than monolithic  predecessors, today’s DevOps teams are still need to upskill on the new paradigm. Once in place, these multicloud apps require new styles of maintenance as well.

Basically, multicloud adds new complexity to the integration task. Let’s count the ways:

1.  multicloud presents DevOps teams with a wider variety of tools, platforms and Kubernetes container orchestration strategies to learn. Added complexity of multicloud native development augers use of automated deployments. Moreover, moving programmed workloads from corporate data centers to single or multiple clouds likely requires refactoring of existing code.  Software architects may argue that moving an application to multiple clouds is a fool’s errand.
2. Application integrations that crisscross clouds and corporate fiefdoms may call for new approaches to DevOps and broader organizational workflows. Better collaboration is needed to remove the Wild West atmosphere from cloud application deployment that often begins with a single developer using a credit card, and then becomes a wide corporate mandate.
3. The counsel to “shift-left” and put security upfront in development has a special flavor in multicloud. Security data collected by different cloud providers takes varied forms; correlating that data can represent a new hurdle. Users are responsible to ensure the security of data traveling from cloud to cloud, too.
4. When it comes to data integration, multicloud implementers are confronted with different data formats and structures. Even seemingly slight differences between different cloud object data formats need attention.
5. Networking in multicloud environments can require special awareness of length of data transmissions, workloads  and specific applications’ sensitivity to latency. Each cloud has its special network security requirements, own set of global regions, and unique connectivity particulars.
6.  Observability of multicloud applications in action requires a sometimes-difficult layer of diverse monitoring software. Rolling up such data to uncover useful performance analytics takes special attention.
7. Financial operations (FinOps) practices are emerging, as users cope with multiclouds’ unique capability to run up bills. Other management issues that take on a new tenor in multicloud implementations are data protection, compliance, risk and governance.

The above challenges. of course, shouldn't deter organizations from pursuing a multicloud strategy given the benefits of insurance against downtime, increased flexibility and cost efficiencies. However, a certain caution is call for,  if only because multicloud has clearly become a multi-year quest.

50 shades of multicloud

David Linthicum, the long-time industry watcher and chief cloud strategy officer at Deloitte Consulting, recently published  “An Insider’s Guide to Cloud Computing” (Addison-Wesley, 2023), which covers the cloud and its multicloud variant. Linthicum, for example, takes on the notion  that multicloud is an antidote to lock-in.

“You don’t really avoid lock-in. If you use a cloud provider’s native features to localize applications and databases, then moving to another cloud means the code and databases still need to be modified to accommodate another cloud provider’s native cloud services. You’re just moving from one walled garden to another. You must adapt application processing, governance, security, database access and so on, to accommodate the ‘new’ public cloud provider’s services.”

There are shades of gray in multicloud lock-in and in portability. With containers, which offer a least-common-denominator approach to development, Linthicum continues, there is room for a greater degree of portability.

But finishing work is required. And no two organizations will approach the task in the exact same way, Linthicum told SDxCentral.

“People should understand that multicloud is going to drive more operational complexity that they have to manage,” he said. “And the byproduct of that is going to be more cost.”

“But the real reason and the value of multicloud is to leverage whatever technology you need to bring more value back to the business,” he said.

The  ability to select among cloud providers allows developers, Linthicum said, “to leverage any database they need, any  AI stack they need, or any governance or security system that they need.”  The goal is to build the best application, capable of bringing the most value back to the business.

Container costs and multicloud tradeoffs

Elements that combine to make multicloud integration possible include enabling technologies that take time to master. Examples are software containers, which hold libraries and runtimes associated with services. In a way, the container that virtualizes operating system functionality carries forward the “run-anywhere” ethos developers have pursued since at least the first days of Java.

Linthicum emphasizes that containers require tradeoffs. They incur upfront effort as well as maintenance once in place. Federated container orchestrations that span multiple clouds, while they may indicate a useful design direction, are not something he recommends.

“The primary driver is best of breed and the reality is we should support developers by allowing them to leverage whatever technology they need to be successful -- to add value to the business by being more innovative,” he said. In this way, Linthicum pictures multicloud as a natural evolution, as it organically adopts innovative technologies from assorted cloud providers.

Cloud and multicloud design decisions

Industry veteran Eric Newcomer said early attempts to follow the multicloud path were challenged by incompatible tooling. As teams tried to move workloads from one to another cloud, there were setbacks.

These setbacks were especially acute around the complexity encountered when integrating the different security, monitoring, logging and other services employed by different cloud providers. Even using tooling certified to run across a variety of clouds did not ensure ease of integration, Newcomer, now principal analyst and CTO at advisory firm Intellyx, told SDxCentral.

Since the early days,he said, architectural decisions on multicloud integration pit best-of-breed services and ‘lock-in’ versus the generic approaches that may be cumbersome to develop and maintain.

“Today, you continue to have this whole debate about going with the native version and getting locked in versus trying to go with something generic and protect yourself from lock in,” Newcomer told SDxCentral.

Debates also continue to revolve around the refactoring choices that need to be made if monolithic applications are going to move to cloud at all. Such applications, as Newcomer described in a recent  blog post on cloud infrastructure, were built  to serve during a traditional business workday of yore – not to serve in an always-on container-based microservices style environment. Moreover, the legacy apps are often built for carefully maintained custom hardware, rather than the always-ready-to-fail-over commodity hardware of the cloud architecture.

Newcomer said a recent development of note is the rise of several independent software vendors in the cloud realm. The list, he said, might range from players such  as database maker Cockroach Labs to ones like cloud data analytics powerhouse Snowflake and others.  These companies can further multicloud by building products that are the same across multiple clouds. In this they stand apart from big cloud providers that are competing on differentiation of their services, and may have less interest in portability and compatibility.

Still a way to go

Some early ideas of multicloud – ones that took the name of cloud brokering and were intended to move workloads around just like commodities traded on a spot market -- have not fully panned out yet, according to Bill Martorelli, analyst, Forrester Research. “The portability aspect really wasn’t there,” he said.

Yet, Martorelli emphasized, multicloud has made progress, but he also cautioned organizations should prepare for operational complexity as they embrace multicloud approaches in their cloud strategies.

The adoption of multiclouds has been organic, accidental (via acquisitions) or incidental over time. But Martorelli sees a growing sophistication.

“One could argue [multicloud] is getting closer now with Kubernetes. But  we still have a way to go. For any customer it’s a question of whether they want to rely primarily on one cloud provider, or whether they want to differentiate based on how they embrace innovation,” Martorelli said. And innovation runs alongside speed to market.

“A few weeks can be a long time in the world we're in now,” said Martorelli, nodding to cloud providers’ recent rapid ramp-up of generative AI capabilities.

The need for speed and flexibility, to be able to take advantage of any new thing the big cloud players come up with, is enabled by supporting more than one cloud.