According to the 2022 Flexera State of the Cloud Report, 89% of cloud customers now use multiple cloud hosting providers. Nearly nine out of 10 might sound high, but it tracks with the anecdotal evidence we’ve seen. For varied and good reasons, companies are spreading their usage across multiple hyperscalers rather than putting all their eggs in one basket.
In this article, I’ll take a look at what life in the cloud(s) looks like today. Along the way, I’ll examine the value that companies and their DevOps teams can derive from this approach, the challenges to be aware of, use cases for making the multi-cloud move and more.
What is Multi-Cloud?
Multi-cloud is a type of cloud strategy approach where an organization uses computing services from at least two cloud providers to run their applications. Instead of a single-cloud stack, these environments include two or more public or private clouds, or a combination of both. With multi-cloud, you can have an application running on Google Cloud Platform (GCP), another in Amazon Web Services (AWS), and maybe one on premise for hard-to-move applications.
There are three kinds of cloud setups. Single cloud is when workloads run solely on one environment. Hybrid cloud entails overlapping workloads across multiple clouds, such as operating the same programs on different providers for business continuity purposes. The final setup is planned multi-cloud, which disperses applications and usage across providers along specific lines, like having compute intensive applications with one provider and data analysis applications with another.
Got a Preference?
Make no mistake, using multiple cloud providers is more complex than a single host, but as its popularity shows, the benefits can carry more weight. Foremost, that’s because an organization gains the flexibility to create a best of all worlds situation, based on factors like the following:
- Operational preference: Users may like different features from different providers, and splitting types of use can yield the ideal setup. For instance, a customer may run compute on AWS but want their business intelligence and machine learning on GCP. In this deployment, the front-end app could gather the needed information, then send the data to the other provider for analysis and ML.
- Regulatory requirements: Certain workloads may require data storage in specific regions, but not all cloud vendors have multiple “regions” in each country. Companies that want to use the cloud can set up a “multi-region” environment by having one region on AWS and a second region on GCP within the same country.
- Cost savings: Cloud providers offer different savings plans. For a savvy user, structuring a flexible program can take advantage of each to significantly reduce costs. It’s also a great way to avoid the pitfalls of lock-in and can make it easier to go elsewhere should a provider lose your trust.
- Failsafe Options: Outages happen, so companies needing 24/7 access to apps will sometimes run portions redundantly on two hosts to ensure smooth operations. For example, a retail app may run on GCP and AWS with live failover between them. That way, if one goes down the other can kick in without noticeable downtime.
Not Without its Challenges
The benefits to running a multi-cloud setup are clear, but there are some challenges as well. A carefully planned multi-cloud setup can save money, but a careless approach will duplicate operations and cost. It also leaves companies vulnerable to mistakes that can lead to significant overrun, especially for hybrid cloud operations, which can be extremely complex. Specialists must be well versed in FinOps principles and experienced at writing modern applications that can scale and work across regions and clouds, to ensure success in a multicloud approach.
Coming to Terms
You’ll undoubtedly come across a lot of new ideas and terminology when researching cloud setups. Here’s a few of the more crucial ones to become familiar with as you plan strategy.
- Lift and shift: This refers to cloud-to-cloud migration, in other words, taking what you have in one cloud and putting it in another, whether it is private to public or public to public.
- Move and improve: This usually refers to migrating while making a small change to take advantage of a public cloud feature. For instance, you could be using open source Kubernetes on premise but want to move to a managed version. Eliminating management of Kubernetes would require a minor change but it improves operations greatly.
- Cloudbursting: On premise offers hard limits to capacity. For companies able to anticipate business peaks, cloudbursting can scale workloads to an on-demand public cloud for a short period. For example, a retailer who operates on premise may cloudburst on Black Friday in order to handle a volume increase.
- High Availability: The goal of this is to eliminate single points of failure by enabling users to continue working even if one of the IT components that an applications depends upon fails.
- Disaster Recovery: Clouds offer solutions to help an organization restore critical systems and data after a disaster strikes. Recovery comes in three “heat levels”: Cold is for data backup only, warm restores operations in several hours, hot is for near-instant resumption of cloud operations.
People and Processes
The most important considerations for operationalizing multi-cloud are people and processes. Providers pump out new services fast, and not only can it be difficult for staff to keep up, it can be demoralizing when they’re unable to. Allowing time to succeed, fail and learn is key, and larger companies might consider designated specialists for each platform.
When it comes to processes, management and security are highly customizable, but it requires deep knowledge to get them right. Operating via infrastructure as code rather than ClickOps is a must. The latter is a time consuming approach that requires selecting options in a provider menu and introduces both operational and security risks in the form of configuration drift.
Life in the Clouds
Multi-cloud brings strong benefits to DevOps. Organizations can get the right infrastructure balance based on their needs. Distributing workloads can reduce the impact of outages and bolster resiliency. There’s greater opportunity for cost savings, and scaling is simple. Ultimately, multi-cloud operational enhancements can result in more reliable and innovative solutions, which delivers on the mantra of DevOps itself.