Cost still biggest driver for multicloud, study finds

Italian insurer Reale Group found itself with four cloud providers running around 15% of its workloads, and no clear strategy to manage them. “It was not a result we were seeking, it was the result of reality,” said Marco Barioni, CEO of Reale ITES, the company’s internal IT engineering services unit.

Since then, Barioni has taken control of the situation, putting into action a multi-year plan to move over half of Reale Group’s core applications and services to just two public clouds in a quest for cost optimization and innovation.

Multicloud environments like Reale Group’s are already the norm for 98% of infrastructure-as-a-service or platform-as-a-service users — although not all of them are taking control of their situation the same way Barioni is.

That’s according to a new study of enterprise cloud usage by 451 Research, which also looked at what enterprises are running across multiple public clouds, and how they measure strategy success.

Two-thirds of those surveyed are using services from two or three public cloud providers, while 31% are customers of four or more cloud providers. Only 2% had a single cloud provider.

Those enterprises’ cloud environments became even more complex when taking into account their use of software-as-a-service offerings. Half of those surveyed used two to four SaaS providers, one-third used five to nine providers, and one-eighth used 10 or more. Only 4% said they used a single SaaS solution, no mean feat given the prevalence of Salesforce, Zoom, and online productivity suites such as Microsoft 365 or Google Workspace.

The study, commissioned by Oracle, looked at the activities of 1,500 enterprises around the world using IaaS or PaaS offerings, or planning to do so within the next six months. The research was conducted between July and September 2022.

Three years on from the first COVID-19 lockdowns, it’s clear the pandemic was a significant driver of multicloud adoption for 91% of those surveyed. But now that the immediate necessity of the switch to remote operations and remote management has passed, enterprises are seeking other benefits as they build their multicloud environments.

Why build a multicloud infrastructure?

The two most frequently cited motivations for using multiple cloud providers were data sovereignty or locality (cited by 41% of respondents) and cost optimization (40%). Enterprises in financial services, insurance, and healthcare were most concerned about where their data is stored, while cost was the biggest factor for those in real estate, manufacturing, energy, and technology.

Next came three related concerns: business agility and innovation (30%); best-of-breed cloud services and applications (25%); and cloud vendor lock-in concerns (25%). Going with a single cloud provider could prevent enterprises from accessing new technology capabilities (such as the much-hyped ChatGPT, which Microsoft is using to draw customers to its Azure cloud services), leave them with a second-best service from a cloud provider less invested in a given technology, or allow the provider to hold them hostage and raise prices.

Traditional benefits of duplicating IT infrastructure were least important, with greater resiliency or performance cited by 23% of respondents, and redundancy or disaster recovery capabilities by just 21%.

But there are still many factors holding back multicloud adoption in the enterprise. Cloud provider management was the most frequently cited (by 34% of respondents), followed by interconnectivity (30%). It was a tie for third place, with data governance issues, workload and data portability, regulatory compliance, and ensuring security across public clouds all cited by 24%.

“The degree to which benefits outweigh challenges may depend on whether multicloud is part of a broader IT transformation strategy … or the extent to which it addresses particular cost, organizational or governance concerns,” wrote Melanie Posey, author of the study. Simply having multiple public cloud environments to meet different users’ needs may be good enough for risk mitigation and cost arbitrage for some enterprises, she wrote, while others will want integrated environments in which workloads and data can run across multiple public clouds.

Reality bytes

Reale Groupe is still straddling those two states as IT leader Barioni moves the company from relationships with four hyperscalers that just happened toward a greater reliance on two that he chose.

His choice of clouds — Oracle’s OCI and Microsoft’s Azure — was constrained by Reale’s reliance on Oracle’s Exadata platform. “Our core applications all run on Oracle databases,” he said.

While several cloud providers offered the packaged services for machine learning and advanced process management he was looking for, the choice of Microsoft to host the remaining business applications came down to latency, he said. Oracle and Microsoft have closely integrated their infrastructure in the regions most important to Reale, allowing the company to build high-speed interconnects between applications running in each cloud. Reale will move its first integrated applications to the cloud in March 2023, he said.

Multicloud management

Johnson Controls is further along in its multicloud journey. It makes control systems for managing industrial processes and smart buildings, some of which can be managed from the cloud-based OpenBlue Platform run by CTO Vijay Sankaran. He said that, while the company has a primary cloud provider, it has chosen to architect its platform to operate across multiple clouds so it can meet its customers where they are.

That multicloud move has meant extra work, connecting everything to a common observability platform, and ensuring all security events feed up to a single, integrated virtual security operations center so that the various clouds can be monitored from a single pane of glass, he said. While the overhead of adding more cloud providers is to be expected, the same problem exists even when dealing with a single hyperscaler, as different regional instances may have specific controls that need to be put in place, he added.

The study also asked enterprises what key outcomes they expected from a multicloud management platform. Only 22% cited the single pane of glass that Sankaran relies on. The top responses were cloud cost optimization (33%), a common governance policy across clouds and integration with on-premises infrastructure (both 27%), improved visibility and analytics (26%), and integration with existing toolsets (25%).

Cost control

Whether an enterprise chooses to spread its workloads across more public clouds or concentrate them on fewer, it all seems to come back to managing cost.

Reale Group’s Barioni has a plan for that involving a core team with a mix of competencies: some technology infrastructure experts, and some with a deep knowledge of accounting. Developers tend to aim for the best technical solution, which is often not the most cost-efficient one, he said.

When applications run on premises, computing capacity — and therefore cost — is limited by what the data center can hold, whereas there are few limits on the computing capacity of the cloud — or its cost. Bringing together the technically minded and financially minded will help Barioni balance cost and performance in this new, unconstrained environment. “Every day, you have to take decisions on prioritizing your workloads and deciding how to optimize the computing power you have,” he said. “It’s a completely new mindset.”