Cloud computing adoption is well established in enterprises, SMEs and startups. But how much cloud should a business adopt? How should workloads be deployed -- across public, private and hybrid clouds? And if multiple cloud providers (public and/or private) are used, which ones should you choose, and how can they be managed to a business's best advantage?
This article examines some of these questions, while the remainder of this ZDNet/TechRepublic special report covers these and other multicloud-related issues in more detail.
It's worth distinguishing between 'multicloud' and 'hybrid cloud' at this point, as there is some scope for confusion here. 'Hybrid cloud' has traditionally meant the combination of private (either on-premises or hosted in a colocation facility) and public cloud infrastructure, with orchestration tools used to deploy workloads and manage the balance between the two -- employing public cloud resources for regular or episodic bursts of compute and/or storage requirements, for example.
'Multicloud', by contrast, has more of a strategic emphasis, describing how enterprises use multiple cloud providers to meet different technical or business requirements. At its most granular, multicloud means cloud-native applications built from containers and microservices using component services from different cloud providers.
How do multicloud strategies play out in practice?
A July 2018 survey by analyst firm Forrester on behalf of Virtustream found that 86% of respondents (727 cloud strategy and application management decision makers in the US, EMEA and APAC) characterised their organisations' cloud strategy as 'multicloud', identifying most with the description 'Using multiple public and private clouds for different application workloads':
Respondents to the Forrester/Virtustream survey defined multicloud in several ways, including: leveraging multiple cloud technologies at once (32%); using public cloud in parallel with traditional non-cloud systems (23%); and using multiple public clouds simultaneously for different workloads (14%).
According to the annual RightScale State of the Cloud Report, the use of multiple clouds is by far the most common pattern among enterprises, with single public, single private and 'no plans' reported by around 10% of survey respondents or fewer. Among enterprises that use multiple clouds, the hybrid cloud model is adopted by nearly 60%, with multiple public and multiple private clouds much less popular (<20%). Over recent years, the number of cloud providers used has averaged just over three for public cloud and just under four for private cloud:
It would seem that the hybrid cloud model, and the use of multiple cloud providers, is the norm among enterprises.
RightScale's findings are broadly echoed by a January 2019 survey from cloud analytics firm Kentik, which reported almost universal (97%) use of AWS (no surprise, since the survey was conducted at Amazon's re:Invent user conference). Along with AWS, Azure was used by 35% of respondents, and Google Cloud Platform by 24%. Company-owned data centres (29%) and colocation/3rd-party data centres (19%) represented more traditional IT environments:
Drilling into the data, Kentik found that a majority (58%) of respondents adopted a multi-cloud strategy -- either using at least two public cloud providers (40%) or all of the 'big-three' providers (18%). Only a third (33%) were hybrid cloud users with company-owned or colo/third-party data centres as well as at least one cloud provider. Kentik ascribed this finding to 'younger organizations whose infrastructure has been cloud-only since their inception'.
Why choose multiple clouds?
It's clear from these and other surveys that organisations routinely use multiple cloud providers. So why are they adopting this strategy?
Vendor lock-in One of the most widely cited drivers of multi-cloud adoption is the desire to avoid becoming locked into a particular cloud provider's infrastructure, add-on services and pricing model. Cloud-native applications based on containers and microservices can certainly be designed to be portable between clouds, but providers will generally try and make their platforms 'sticky' with specific functions and services that differentiate them from their rivals. This means that a portable 'lowest common denominator' application may not exploit a cloud provider's full potential, and that businesses will therefore have to determine the trade-off between portability and full functionality -- with potential lock-in -- for particular workloads. The result across multiple workloads is likely to be a multicloud strategy.
As Gartner analyst Michael Warrilow puts it: "Most organizations adopt a multicloud strategy out of a desire to avoid vendor lock-in or to take advantage of best-of-breed solutions... We expect that most large organizations will continue to willfully pursue this approach."
Shadow IT An organisation may end up with a multicloud strategy by accident, via the agency of 'shadow IT' -- that is, technology adopted by business units independently of the IT department, which may subsequently be 'reined in' for oversight by the CIO.
The extent of shadow IT revealed by McAfee's 2019 Cloud Adoption and Risk Report is startling: 1,400 IT professionals in 11 countries were asked to estimate the total number of cloud services in use in their organisation and came up with an average of 31. The actual average figure was 1,935:
Here's how those (actual) cloud services break down by category:
As you'd expect, the full gamut -- SaaS, IaaS, PaaS and other as-a-service variants -- is present.
Performance Organisations can minimise latency -- and other performance metrics, such as jitter and packet loss -- by choosing a cloud provider with data centres that are geographically close to their customers, as performance is generally inversely correlated with the number of network hops between servers. For enterprises with a wide range of cloud-based workloads, the optimal solution is likely to involve multiple cloud providers.
Compliance Data governance requirements -- such as the EU's GDPR, for example -- will often require customer data to be held in particular locations. Unless organisations are willing to create and maintain their own on-premises data lakes, this will often require a multicloud approach -- again, depending on an enterprise's geographical distribution and workload mix.
Resilience All cloud providers -- even hyperscale ones with multiple geographically dispersed, redundant data centres -- suffer outages from time to time, so putting all your workload 'eggs' in one provider's 'basket' runs the risk of a mission-critical application becoming unavailable. A multicloud strategy may bring deployment and management headaches (see below), but it should also make for better security, failover and disaster recovery -- in a word, resilience.
AWS, Microsoft Azure and Google Cloud Platform all provide information on outages, but reporting differences make comparisons tricky. Analyst Zeus Kerravala recently had a go, and found that, through 2018 to May 2019, AWS and GCP had comparable levels of IaaS downtime (338h and 361h respectively), but Azure was in a different league with 1,934 hours -- over 5 times the outage rate for AWS and GCP.
Kerravala concluded: "I'm certainly not saying to not buy Microsoft Azure, but it is important to do your homework to understand the historical performance of the services you're considering in the regions you need them. The information on the vendor websites may not tell the full picture, so it's important to do the necessary due diligence to ensure you understand what you're buying before you buy it."
In Spiceworks' recent Public Cloud Trends in 2019 and Beyond survey, the Big Three cloud providers were ranked similarly on 'Maximum uptime', with bigger differences evident in 'Simple to manage', 'Optimal data center location' and 'Hyperscalable offerings':
Managing a multicloud deployment
A multicloud strategy has many advantages, but it undoubtedly adds an extra layer of management complexity -- especially if multicloud adoption develops in an ad hoc manner rather than being planned from the ground up. In the Forrester/Virtustream survey referenced earlier, there was a roughly equal split on the latter issue:
Given the level of ad hoc adoption, it's no surprise to find that respondents favoured cloud providers offering add-ons such as cloud strategy service, deployment, migration and management capabilities. They also preferred vendors with whom they have long-term IT partnerships, and tend to reach out to a pre-approved list of vendors.
In Spiceworks' Public Cloud Trends in 2019 and Beyond survey, 'Managing multiple cloud solutions' ranked fifth in the areas where businesses need more support from cloud vendors, with this requirement more pressing for small and mid-size businesses than enterprises:
There's an increasing number of solutions available to help organisations manage multiple cloud deployments.
Check out Gartner's first Magic Quadrant for Cloud Management Platforms (CMPs), which "enable organizations to manage multicloud (i.e., private and public cloud) services and resources". Functional areas for CMPs include: provisioning and orchestration; service request management; inventory and classification; monitoring and analytics; cost management and resource optimisation; cloud migration, backup and disaster recovery; and identity, security and compliance. Leaders in this space -- according to Gartner -- are Flexera (RightScale), Scalr, Embotics and Morpheus Data.
Gartner's latest Magic Quadrant for Public Cloud Infrastructure Professional and Managed Services, Worldwide also has a multicloud focus, noting that "Customers now have requirements to manage their multiple clouds from a unified perspective, and MSPs [Managed Service Providers] have been working to deliver tooling, processes and staffing that are a blend of cloud-agnostic and cloud-specific capabilities." However, despite requiring "Proven deep and broad expertise with AWS, Microsoft Azure, Google Cloud, Alibaba Cloud, IBM Cloud and Oracle Cloud," Gartner judges that "no MSP has built a completely transparent, fully integrated multicloud solution".
An interesting recent development is Google's Anthos, a Kubernetes-based suite of GCP components that will not only allow organisations to run applications on-premises and in Google's public cloud, but also to "manage workloads running on third-party clouds like AWS and Azure, giving you the freedom to deploy, run and manage your applications on the cloud of your choice, without requiring administrators and developers to learn different environments and APIs." This is a potentially disruptive move in the rapidly evolving cloud-native hybrid/multicloud space.
Modern cloud-ready applications are built from containers and microservices (and are increasingly becoming further abstracted to functions-as-a-service). This makes them suitable for deployment in private and/or public clouds from multiple vendors, in order to achieve the optimal combination of performance, cost and other key metrics.
Orchestration tools are beginning to catch up with the multicloud movement, making it easier for businesses to manage their workloads in this environment, but this is a fast-developing market.