Cloud repatriation vs. multi-cloud: IT seeks cost relief
Amid fiscal worries, IT orgs reassess cloud investments. Whether they go back on-premises or expand to multi-cloud, they face similar distributed computing management challenges.
With talk of an economic downturn looming after three years of pandemic-accelerated digital transformation, enterprises are coming to a fork in the road with cloud costs.
What had been a race to adopt cloud computing and cloud-native tech as COVID-19 forced businesses to go digital has given way to belt-tightening amid inflation, rising energy prices and predictions of a coming recession. FinOps features are finding their way into IT ops tools as cloud costs become a growing concern, and some enterprises are undergoing a deeper rethink about their cloud strategy in general.
Some IT organizations are choosing to move resources back into private data centers for cost control and predictability, a concept termed cloud repatriation. Others are forging ahead with public cloud, but seek pricing leverage among competing service providers through a multi-cloud strategy. And in either case, a shifting approach to cloud lends itself to more broadly distributed computing infrastructures that require fresh management tactics from IT teams -- a need that major IT vendors are rushing to fill.
"I am in many conversations with vendors about solving these challenges with harmonization of the tech stack," said Paul Nashawaty, an analyst at TechTarget's Enterprise Strategy Group. "It is complicated and not seamless."
Cloud repatriation yields quick ROI
Park 'N Fly, a travel services company in Atlanta, has moved half of its public cloud resources back into self-hosted data centers in the last year. This move was spurred in part by the fact that Microsoft Azure prices haven't fallen the way IT leaders expected with a general increase in enterprise cloud adoption. In fact, recent macroeconomic trends such as rising energy prices and inflation have been passed on to cloud customers by Azure in certain areas. AWS didn't consistently lower prices over the previous three years either, according to some market research studies.
Ken SchirrmacherCTO and senior director of IT, Park 'N Fly
Some services such as Microsoft 365 also increased in price last year, while by-the-hour infrastructure costs piled up with Park 'N Fly's business growth. Plus, the ease of spinning up cloud assets and a propensity for users to forget about them helped costs balloon.
"We always knew that the cloud was more expensive than an on-premises [architecture]," said Ken Schirrmacher, CTO and senior director of IT at Park 'N Fly. "The hope was that if there's a larger [user base], the services would get cheaper and cheaper. What we've seen is the exact opposite."
Buying data center colocation space and hardware also represents a substantial investment, but it came at a time when the IT infrastructure needed a refresh whether it stayed in the cloud or not, Schirrmacher said.
"Even with all the money that we spent re-buying all that server infrastructure, our return on investment came four months out," he said. "That's how expensive the cloud is."
The company might move another 25% of its IT assets out of the public cloud, Schirrmacher said, but some, such as email services, will remain. Park 'N Fly will also use public cloud resources to back up on-premises apps for failover in the event of an outage and will back up data from cloud services to on-premises.
To maintain cohesion among its hybrid cloud assets, Park 'N Fly is moving back to Octopus Deploy for CI/CD pipelines from Azure DevOps. Octopus, which Park 'N Fly retired in its initial public cloud migration, is back in favor because it can do what Schirrmacher calls "one-push wonders" to cloud and on-premises infrastructure at the same time. Other integration points such as Cisco's Meraki SD-WAN and Kubernetes container orchestration help tie together cloud and on-premises architectures.
Moving from all in on public cloud to self-hosting half of its IT services over the last year followed the same pattern as the company's hybrid approach to remote work, Schirrmacher said.
"We just designate a time where we're all going to be in the office, so that people can travel and focus on their roles the other days, but you still get that camaraderie and you still can see the people that you're working with," he said. Whether employees or IT resources, "if you put all your eggs in one basket, that's typically not good," he added. "Doing a little bit of everything is the best mix."
Multi-cloud empowers, tied in with observability
Insurance provider Employers Holdings went all in on public cloud in a matter of nine months in 2018 and plans to remain there. However, from the beginning, the 110-year-old company emphasized a multi-cloud approach, with assets spread over public clouds including AWS, IBM Cloud, Oracle Cloud and others.
Employers does its best to avoid tying into any one cloud vendor's services, a strategy its IT leaders said makes it easier to take its business elsewhere should they sour on a provider -- or its pricing.
So far, Employers, which has thousands of VM instances in AWS alone, has been able to keep cloud costs flat year over year through renegotiation with cloud vendors, said Jeffrey Shaw, executive vice president and CIO at the insurance company headquartered in Henderson, Nev.
"I thought what would happen is we'd get into AWS and then they'd just launch into us for money," Shaw said. "But they haven't -- they've been pretty consistent with the pricing, and we just did a renegotiation where we got more discounts than we got the previous time."
It also takes continuous effort, Shaw acknowledged, to keep cloud costs under control, but there are effective cloud cost management tools that allow for this.
"We spend a lot of time and effort culling our cloud [resources] to keep runaway [spending] from happening," he said. "If we didn't do that, we would have tripled the costs of our cloud over the last five years."
In addition to renegotiating prices and monitoring cloud spending, Employers is also selectively using a few cloud providers' value-add services, such as Oracle Cloud's support for Oracle databases, but always keeping its options open, said Jeremy Harkins, the company's cloud architect and director of IT.
"It isn't just a lift and shift and forget -- you have to stay on top of what all the providers are doing, your use cases, and can you move [an application] somewhere else and gain more capability while reducing cost?" he said. "We look for all these win-win options."
For now, what ties Employers' multi-cloud deployments together is primarily observability tools from Dynatrace, along with Ansible infrastructure as code and Kubernetes container orchestration for most apps, as well as VMware Cloud for VMs.
But the company also uses Red Hat OpenShift Kubernetes for some apps, and a variety of DevOps pipeline tools such as Jenkins and Bitbucket, while it's evaluating Tekton event-driven CI/CD pipelines. As with most multi-cloud users, Employers doesn't move apps between clouds on a regular basis yet and is still exploring how to create a single point of policy management for all its cloud resources, Shaw said.
"We've been asking VMware or AWS or someone to come up with a cross-cloud platform, and they all say they have them, but we have not found one that actually works correctly," especially with niche cloud services from IBM and Oracle, Shaw said. "VMware is best positioned with their hypervisor to move across clouds -- not just with containers, but with VMs -- but they haven't quite been able to crack that yet."
VMware, fresh from a $2 billion public commitment to its R&D from parent-company-to-be Broadcom this month, has its sights set on cross-cloud management as its primary product strategy, according to briefings at a recent industry analyst day.
Employers is also a customer of PagerDuty, which launched a new runbook-based IT automation product last month, Operations Cloud, that offers policy-based management across clouds and devices ranging from IoT sensors to servers. It plans to investigate that as an option, according to Harkins.
"It looks compelling, [with] a full lifecycle of capabilities covering software builds, deployment, monitoring and automated self-service remediation for customers throughout the business," he said. "We have PagerDuty in our environment, and it hasn't been used to the fullest extent of its capabilities. However, we do intend to make it a central component of our alerting and incident management."
Beth Pariseau, senior news writer at TechTarget Editorial, is an award-winning veteran of IT journalism. She can be reached at [email protected] or on Twitter @PariseauTT.