Over a decade after the launch of Amazon Web Services, AWS continues to benefit from the “first mover” position. While AWS is still the market leader, cloud services from other industry titans such as Google or Microsoft have grown in popularity (as have specialized services from other vendors), giving enterprises unparalleled choice in the cloud services they use.
As a result, enterprises and developers increasingly consume cloud services from a range of different vendors, giving rise to the term “multicloud.” Not all multicloud is created equal, however.
Some multicloud deployments are intentional in nature, driven by a strategic plan to improve business agility or to tap into best-of-breed services. Other multicloud deployments are somewhat accidental in nature and resemble traditional enterprise IT, fraught with shadow IT complicating CIOs’ attempts to bring discipline to their data center.
TechRepublic’s cheat sheet to multicloud is an introduction to using multiple cloud providers, with an emphasis on intentional multicloud. This guide will be updated as new integrations and services become available. It is also available as a download, Multicloud: A cheat sheet (free PDF).
Multicloud refers to the practice of using services from multiple heterogeneous cloud service providers, including AWS, Google Cloud Platform or Microsoft Azure as well as specialized PaaS, IaaS or SaaS providers. Multicloud is not the same as hybrid cloud, which blends the use of private cloud environments and public cloud environments.
Multicloud as a strategy should not be confused with the simple fact of using cloud services from different providers, which is how enterprise IT has always functioned.
As an architectural choice, multicloud can be used for a variety of reasons. The most obvious one is disaster recovery. While cloud vendors offer a variety of options and SLAs (service-level agreements) for redundancy to guarantee uptime and backups to ensure data integrity, both of these rely on the supposition that the vendor’s entire infrastructure does not fail at once.
However, for most enterprises most of the time, organizations should ensure they’ve architected for resilience across a single cloud provider’s regions before endeavoring the much more difficult task of architecting multicloud resilience across different cloud providers.
While most workloads can be built to be vendor neutral (this flexibility is a primary benefit of multicloud), many workloads benefit from using specific cloud platforms. As analyst firm Thoughtworks has suggested, “Opting for an always-on, cloud-agnostic architecture is costly and effort-intensive,” while embracing the innovations in a particular cloud enables developers to be more productive.
As such, it pays to embrace different clouds for their particular strengths. For example, apps that use Alexa Skills are better served by using Amazon Web Services, as the APIs involved are native to AWS. Likewise, supported languages and depth of ability for natural language processing varies widely between different cloud providers.
Roughly one third of the IT professionals surveyed in TechRepublic Premium’s Managing the multicloud survey indicated their organization uses a specialized application or solutions provider, such as Google Drive, Salesforce or Cloudflare.
These are closer to services than they are cloud platforms. While there is feature duplication between these and similar companies as with public cloud, these products do not support general compute workloads commonly associated with cloud computing.
Multicloud’s main advantage is that organizations and application developers can pick and choose components from multiple vendors and use the best fit for their intended purpose. To draw a comparison, multicloud is more à la carte than table d’hôte.
However, this benefit may depend upon a microservices architecture where an organization could, for example, host its live e-commerce site with customer data and product catalog on AWS and then have a replica hosted on Google Cloud, which is creating personalization and offers from customer interactions.
Which services an organization chooses will depend upon existing IT systems and the skillsets of its employees. For organizations with an outsized dependency on the Windows ecosystem, for example, leveraging some Microsoft Azure services may be beneficial, while the same organization may use Google Cloud for machine learning and analytics and/or Amazon for public-facing web services.
Industry cloud service providers, such as Dell Boomi, offer cloud-focused digital transformation and data management services for targeted industries like healthcare, which traditionally face regulatory hurdles to modernization. These hurdles include data storage requirements for regulatory compliance and disaster recovery or business continuity when integrating with public cloud services.
Another reason to consider multicloud is resilience across cloud providers within a single region. As robust as the infrastructures of a Microsoft Azure or Google Cloud, the reality is that systems fail.
To make multicloud resilience more reality than promise, an organization would need to architect the same resilience into both data and application tiers, as there isn’t much benefit to having the data tier available after an outage of one’s primary cloud provider, but not being able to access it from an application tier that hasn’t failed over.
Multicloud is also ideal for data sovereignty and elimination of concentration risk. This has become a growing concern for those within Europe.
Additionally, multicloud offers low latency access to an organization’s application. As cloud demand has increased, organizations have reported capacity constraints across a number of clouds — between low or even no capacity. An organization’s preferred cloud provider may not yet have data centers running in a region close to their customers.
Multicloud allows an organization to mix and match infrastructure to support an application, provided the application has been architected specifically to handle running across multiple clouds. For example, a specific cloud service may be available but in a region that is too geographically distant to produce good application performance, making an alternative cloud provider the smarter choice.
As mentioned, hybrid clouds comprise different kinds of clouds, while multicloud includes multiple providers of the same type of cloud. In other words, a hybrid cloud approach might involve building applications that span a private data center and Google Cloud, whereas a multicloud architecture might combine services from Google Cloud and AWS (both public IaaS providers).
A hybrid cloud strategy will tend to suit organizations that aren’t yet ready to commit all applications to the public cloud. It can be used to help organizations bridge their existing infrastructure, not to mention the skill sets of their employees, to a more cloud-centered future.
This depends. A SaaS provider will require a multicloud approach, but for many enterprises, the complexity involved in successfully managing a multicloud environment may be too difficult. Generally, a multicloud deployment will be useful for organizations that have specific needs or dependencies to satisfy, such as integrations with Internet of Things (IoT) devices or a reliance on Windows software or specific third-party solutions.
Multicloud offers a great deal of flexibility in how resources are managed, though the difficulty increases roughly exponentially with the number of integrations added. Cloud management platforms can be used to ease deployment and integration of various cloud services. There is a growing number of SaaS providers offering managed cloud services for databases.
Keep in mind that there is no magical tool — no proverbial “single pane of glass” — that will make multicloud easy. While it’s now possible to buy fully managed database services, for example, with the SaaS provider taking care of failover across clouds, this is only as useful as the multicloud architecture powering the application tier.
Because there are no tools that can offer a multicloud panacea, any enterprise looking to embrace multicloud needs to first look into training its employees to be proficient with multiple clouds.
A developer proficient with AWS services will not necessarily be effective with Microsoft Azure services, and “commodity” services (like compute and storage) are not truly commodity, given the innovations each cloud provider invests into their products.
The complexity associated with multicloud requires deep expertise on different clouds within one’s organization, which will often prove difficult to find.
Multicloud is continuing to gain popularity as competitors to Amazon Web Services have appeared and particularly as specialized cloud technology vendors and service vendors have gained traction. But today, multicloud is more popular to talk about doing than to actually do.
As organizations grow, it may be the case that needs for individual teams or projects are not met by their existing cloud provider; likewise, for mergers and acquisitions, not all business operations can be easily migrated to the cloud infrastructure of the acquiring company. These are optimal cases for adding a secondary public cloud provider for a multicloud deployment.
A study from TechRepublic found that 81% of organizations reported using multicloud in 2020, up from around two-thirds in 2019. Multicloud deployment plans were affected by the COVID-19 pandemic, with 40% of respondents saying it accelerated deployment plans.
Flexera’s 2020 state of the cloud report found 93% of organizations have plans in place to use multiple cloud providers, which ZDNet’s Larry Dignan said indicates that enterprises are “all in” on multicloud.
Given that most enterprises are still in their infancy with cloud adoption and the inherent complexity of multicloud, such survey data likely reveals the most basic of multicloud adoption (multiple clouds running, but applications not architected to take advantage of multicloud).
Migrating to a multicloud deployment is not a decision that should be entered into lightly, as it can introduce significant complexity. Enterprises that want to reduce that complexity can turn to SaaS providers that take care of the burden of managing databases or other infrastructure across diverse clouds, but need to keep in mind that successful multicloud deployments depend upon multicloud architectures that span application and data tiers.
One way to immediately improve multicloud portability is by building upon common open-source substrates. For example, a company can choose PostgreSQL and use it across public clouds and private data centers.
The key here is to use the most basic version of the open-source project in question to ensure it can run anywhere. Cloud management platforms can be used to avoid potential issues with common configurations; though, some corner cases can hamper successful deployment.
Generally speaking, successful multicloud deployments are still difficult enough to accomplish that organizations are wise to look to experienced outside help.
Google has been the conspicuous leader in multicloud among the public cloud companies. Though Microsoft offers Azure Arc, which enables customers to extend Azure services to private data centers and other cloud providers, Google’s Anthos set the standard and is arguably the industry’s first true multicloud product.
Anthos gives customers the ability to run Kubernetes and other workloads across private data centers and public clouds like AWS and Azure. No hypervisor layer is needed. AWS, for its part, has largely eschewed multicloud; though, it has made its first foray with EKS Anywhere to enable customers to run the company’s EKS service wherever they choose. It’s a start but, again, Microsoft and, especially, Google are significantly ahead with multicloud.
Outside the big cloud providers, there is a lot of multicloud marketing, promising a single pane of glass to centrally manage applications, but the substance behind these claims is generally lacking. Many tools vendors claim to provide a single view of services running on different clouds, but cloud service providers launch new services and variations at such a speed that it is essentially impossible for the tools to keep pace.
The one exception may be Red Hat, as its OpenShift product has long supported deployment to multiple clouds, and Red Hat also has a strong focus on hybrid cloud through OpenShift.
Because of the difficulty of pulling off multicloud management, enterprises interested in running across multiple public cloud providers should look to SaaS providers like MongoDB, Redis Inc. or DataStax, precisely because they only need to focus on managing their application/infrastructure. This is a far more feasible task than purporting to manage every possible permutation of cloud service across diverse infrastructure with wildly varying compute, storage, networking and more.