As organisations embark on their digital transformation journeys, modernising business processes and optimising IT infrastructures, it's clear that cloud computing services, along with automation, orchestration and data analytics, will become increasingly important.
These trends are well underway -- for example, in F5's 2020 State of Application Services Report (February 2020), nearly a third of respondents reported that over half of their applications will be in the cloud by the end of 2020. The ongoing coronavirus pandemic will undoubtedly accelerate this cloud migration process.
SEE: Nextcloud Hub: User tips (free PDF) (TechRepublic
But, as F5's report details, enterprise application portfolios are typically a mixed bag, including four main architectures: three-tier web and mobile (40%), client-server (34%), microservices/cloud-native (15%), and old-school mainframe/monolith (11%). This means that the details of cloud migration will vary depending on an enterprise's particular mix of application types -- both external- and internal-facing.
Drilling down, F5 found that most enterprises (76%) use some combination of traditional (3-tier/client-server/monolith) and modern (mobile/microservices) architectures, while 21% use only traditional and just 3% use only modern; 18% use all five architectures and 11% use just one.
App portfolio heterogeneity is the reason why, when respondents were asked "How does your organisation decide which type of cloud is best for your application?", the top answer was on a 'case by case, per application' basis. It's also why three-quarters of F5's respondents reported that they had applications in two or more cloud providers, and why over a third said that refactoring legacy applications for modern environments is a priority for digital transformation.
This mix of application architectures in a typical organisation's portfolio "highlights the fact that multicloud will be the norm for the long term," F5 says.
This article provides an overview of the issues surrounding multicloud environments, following a similar exercise in 2019. For more detail on the current state of play in this area, check out the remainder of this ZDNet/TechRepublic special report.
In an ideal world, application workloads -- whatever their heritage -- should be able to move seamlessly between, or be shared among, cloud service providers (CSPs), alighting wherever the optimal combination of performance, functionality, cost, security, compliance, availability, resilience, and so on, is to be found -- while avoiding the dreaded 'vendor lock-in'.
"Businesses taking a multicloud approach can cherry-pick the solutions that best meet their business needs as soon as they become available, rather than having to wait for one vendor to catch up," John Abel, technical director, office of the CTO, Google Cloud, told ZDNet. "Avoiding vendor lock in, increased agility, more efficient costs and the promise of each provider's best solutions are all too great to ignore."
That's certainly the view taken by many respondents to the survey underpinning the 2020 State of Multicloud report from application resource management company Turbonomic. In particular, respondents from organisations that see themselves as 'leaders' in terms of the ability to leverage new technologies to advance their business goals were the most positive about multicloud:
The ideal may be seamless cloud-hopping, but there are plenty of real-world challenges involved in migrating applications to the cloud, and managing them in a multicloud environment. In F5's survey, the principal one, by some distance, was Maintaining security, policy and compliance:
Flexera's 2020 State of the Cloud Report states that "Due to its complexity and dynamic nature, the multi-cloud environment brings many challenges, such as assessing the suitability of on-premises apps for migrating to the cloud". Specifically, Understanding app dependencies emerged as the top cloud migration challenge, followed by Assessing technical feasibility and Assessing on-prem vs cloud costs:
When it comes to overall cloud challenges, as opposed to migration, the top four in Flexera's 2020 survey were Security, Managing cloud spend and Lack of resources/expertise. Managing multi-cloud made the top challenges list, but some way further down:
In Turbonomic's 2020 State of Multicloud report, cultural and/or operational factors, and the complexity of managing hybrid/multicloud environments were the foremost challenges faced by organisations:
"Bottom-line, cultural change is a prerequisite for managing the complexity of today's hybrid and multicloud environments. Teams must operate faster, dynamically adapting to shifting market trends to stay competitive. This new reality requires tighter collaboration between teams, as well as a maniacal focus on the customer experience and how it impacts the business," the Turbonomic report said.
Google Cloud's John Abel also identifies complexity as a key barrier when it comes to multicloud: "Perhaps the biggest cause for hesitation around multicloud adoption is the complexity of deploying more than one cloud platform, and it's particularly difficult to do without hindering productivity or innovation. To overcome this, open-source technologies such as Kubernetes can aid the orchestration of containers to limit disruption with each new implementation and enable development teams to efficiently move workloads around."
How many clouds, and which ones?
The use of multiple clouds is by far the most common pattern among enterprises, with 93% adopting this strategy in Flexera's 2020 State of the Cloud Report (87% hybrid cloud, 6% multiple public cloud). This continues a long-standing trend:
Among the 87% of enterprises with hybrid cloud deployments in 2020, the most common combination is multiple public and private clouds (53%), followed by multiple public/one private (33%), one public/multiple private (7%), and one public/one private (7%).
In 2020, Flexera reported that organisations, on average, were using 2.2 public and 2.2 private clouds, and were experimenting with a further 1.2 public and 1.7 private clouds. Here's how these numbers have varied in previous years:
Cloud usage, particularly of the public cloud, has steadily increased since 2016, while experimental deployments have fluctuated, generally at a lower level.
Amazon Web Services (AWS) is still the leading public cloud in Flexera's 2020 survey. However, the rise of Microsoft Azure since 2016 is notable, leaving Google Cloud some distance behind in third place:
Among private cloud providers, VMware vSphere remains the leader, with Microsoft's Azure Stack showing a similar trajectory to its public cloud counterpart and claiming second place in 2020. Other prominent private cloud offerings are the open-source OpenStack and VMware's vCloud Director, while AWS Outposts nearly doubled its adoption rate between 2019 and 2020:
In Turbonomic's 2020 State of Multicloud survey, Microsoft Azure was the leading public cloud (61%, up from 52% in 2019), with AWS relegated to runner-up (53%, down from 55% in 2019). The third most-used cloud type was 'Private cloud self-service and elasticity with containers and/or VMs, on-premises and/or hosted at an MSP' (38%, down from 45% in 2019). Respondents mostly used one (33%) or two (30%) clouds in 2020, with 29% using three or more:
"As organizations (and cloud offerings) become more sophisticated with cloud services, we can expect this graph to shift right," the Turbonomic report said.
What about on-premises?
Although multicloud will become the norm, there's still a place for the on-premises data centre, at least in the near term, either as part of a hybrid cloud strategy or to host legacy applications that, for whatever reason, are not suitable for migration to the cloud.
For example, in the Forrester/IBM report The Key To Enterprise Hybrid Multicloud Strategy (January 2020), a survey of 350 global enterprise IT decision makers found that more than half of mission-critical workloads and 47% of data-intensive workloads will still be run either on-premises or in an internal private cloud in two years.
The main reasons for using on-premises resources to host workloads focus on compliance, concerns about data in transit (cost, latency or security), and application/infrastructure performance:
According to the Forrester/IBM report, organisations also "use on-premises and private cloud to side-step bureaucratic processes and kick-start development efforts".
Concerns about data are clear in Flexera's 2020 State of the Cloud report, which shows that non-sensitive data is far more likely to end up in the cloud than corporate financial or consumer data. Only a quarter (26%) of respondents were prepared to put all or most of their corporate financial data in the cloud, for example, compared to nearly half (46%) who wanted this sensitive information to remain fully or mostly on-premises:
This concern over sensitive data, and the desire to keep it safely on-premises, underpins moves like Google's 'hybrid AI' offerings, such as Speech-to-text On-Prem, which has recently become available on its Anthos platform. "By bringing AI on-prem, customers can now run AI workloads near their data, all while keeping them safe," Google Cloud's John Abel told ZDNet. "Hybrid AI simplifies the development process by providing easy access to best-in-class AI technology on-prem."
Multicloud cost management
A key problem for organisations using multiple clouds is keeping a handle on costs. Cloud pricing is complex and fast-changing, and billing models vary between cloud platforms, yet cloud services can easily be accessed and provisioned -- often by business units acting independently of the IT department. This is a combination that can result in unexpected and uncontrolled costs.
For example, in Flexera's 2020 State of the Cloud report, survey respondents estimated that 30% of their organisations' cloud spend is wasted, while Flexera puts the wastage figure at 35%, based on analyses of its customers' cloud spending.
The major public cloud providers -- Amazon Web Services (AWS), Microsoft Azure and Google Cloud Platform -- all offer native tools to manage costs, and these will suit organisations that limit themselves to one public cloud platform. However, third-party tools could potentially do a better job of multicloud cost management.
Analyst firm Gartner recommends a hybrid approach: IT managers should use the cloud providers' native tools first, identify gaps in functionality, and then address these with third-party offerings. When it comes to third-party tools, Gartner suggests favouring vendors that offer "sophisticated analytics that ensure high-quality insights", and that IT pros should "plan for multicloud, and prioritize solutions with higher feature parity between platforms".
Last year Gartner compared five leading third-party cloud cost management vendors -- Apptio, CloudCheckr, Flexera, Turbonomic and VMware -- and generated the following summary, noting that "the cost optimization market still has many unexplored areas":
In its more recent How to Manage and Optimize Costs of Public Cloud IaaS and PaaS report (March 2020), Gartner says: "Third-party cost management tools provide functionality that can exceed what cloud providers natively implement. Furthermore, their support for multiple cloud platforms allows organizations to implement a multicloud management strategy". The analyst firm adds that the provider-independence of these tools will allow them to continue to receive investments going forward.
Missing from the above analysis is the biggest component of cloud spending worldwide: software-as-a-service, or SaaS. In Gartner's most recent forecast, SaaS spending is expected to rise from $101.48 billion in 2020 to $138.26bn in 2022 -- 36.2% growth over two years. However, IaaS and PaaS will grow faster over the same period, at 59.9% and 57.4% respectively. Total spending on public cloud services is expected to grow by 18.4% to $305bn in 2020/21 and by 18.8% to $362.3bn in 2021/22:
Because it's particularly easy to deploy, SaaS is the most likely category of cloud spend to get out of hand. Fortunately, there are plenty of tools that allow organisations to discover, manage and optimise their SaaS subscriptions. Leading SaaS cost management tools include Zylo, Torii and Augment. Among the broader-based vendors, Flexera offers SaaS Manager, while Apptio has recently added SaaS cost management to its Cloudability offering.
Managing the multicloud
An organisation may report that it's using multiple clouds, but that doesn't necessarily mean its applications are moving freely between different cloud providers. That not only requires cloud-native applications built around containers and microservices, but also multicloud management tools, which are still evolving.
SEE: IT Data Center Green Energy Policy (TechRepublic Premium)
For example, in Flexera's 2020 State of the Cloud survey, the most common multicloud implementation was Apps siloed on different clouds (55%), although data integration and workload mobility between clouds, and apps that span public and private clouds are also represented:
Although the use of containers is widespread, with 65% of organisations using Docker and 58% using Kubernetes (the Docker orchestration tool) in Flexera's 2020 survey, respondents also reported multiple challenges related to their use, headed up by Lack of resources/expertise, Security and Migrating traditional apps to containers:
In Turbonomic's 2020 State of Multicloud survey, respondents reported that 26% of their environments were running containerised applications, rising to 40% within 18 months, and that 28% of containerised apps were 'mission critical'.
Given that advanced multicloud architectures have yet to become widespread, and the challenges surrounding containers, it's not surprising that only around a third of Flexera's respondents are using multicloud management tools:
"Managing different cloud systems can certainly be a challenge, particularly if multicloud adoption develops in an ad hoc manner rather than being planned from the ground up," said Google Cloud's John Abel. "To manage processes and workflows across different cloud services, many organisations look to cloud orchestration software to host and coordinate cloud services from different providers."
Author / sponsor
~2600 worldwide 'across a range of industries, company sizes, and roles'
750 'cloud decision makers and users' worldwide, (554 enterprise, 196 SMB)
Forrester / IBM
350 'global decision makers for IT infrastructure environments'
938 users (50%), director/managers (41%) & executives (9%)
There are many reasons why organisations make use of cloud services from multiple providers -- accessing the best functionality for particular purposes, hedging bets against outages, avoiding vendor lock-in, and more. However, the ease with which cloud-based infrastructure can be provisioned, and software subscriptions taken out, also means that multicloud deployments are likely to evolve semi-independently of the IT department, with potentially negative consequences.
Organisations face many challenges in the quest for free movement of application workloads across different clouds. Prominent among these are: the heterogeneity of enterprise application portfolios, with significant numbers of 'legacy' apps requiring significant refactoring to migrate to the cloud; security concerns; cost control; lack of resources and expertise; the need for cultural changes in IT teams; and management complexity.
But multicloud is undoubtedly here to stay. The next step is to make its more sophisticated incarnations easier to deploy and manage. To this end, Google Cloud's John Abel sees a key role for open-source APIs in enabling future multicloud developments: "2021 will see a continuation of an ever-evolving workplace, and businesses will need to remain agile, responsive, and able to adapt to survive...To enable multicloud deployments, build new environments and modernise old ones, the open-source community will dial up investment in container and serverless functions, creating a spike in global demand."