Cloud-first has become the guiding principle for application modernisation and migration. Cloud-native is the mantra for application development.
Private cloud – typically an on-premises data centre that enterprises are attempting to retrofit with virtualisation, automation, self-service, capacity planning, and other features that come built into the public cloud – has definitely taken a back seat.
For a time, many enterprises seemed to be migrating as many apps as possible to the public cloud and relegating the private cloud to hosting an ever-shrinking collection of legacy, highly customised, out-of-support, end-of-life apps.
But as we enter the second decade of the cloud revolution, a new vision has emerged in which private clouds become an equal partner in an integrated, multi-cloud world of private, public, and edge clouds.
Gartner is calling this the distributed cloud, and Michael Warrilow, research vice president for infrastructure software, predicts that "it has a great potential for success."
Warrilow in 2018 framed the issue facing traditional private clouds, which didn't have the infrastructure to match hyperscale public cloud. "Infrastructure and Operations leaders must resist the temptation to mimic a style of computing that they are ill-equipped to replicate," he wrote in a research report.
But if you can't beat them, join them. IT execs don't need to try to replicate public clouds on their own anymore, because public cloud vendors are now offering to deliver the best features of the public cloud right to the doorstep of on-premises or co-located private cloud environments.
The vision, according to Forrester analyst Chris Gardner, is for enterprises to run workloads on the appropriate platform, whether that's public cloud or on-premises, and to strive to provide private clouds with equivalent functionality to public clouds to the extent possible. No more second-class citizen.
Challenges of the private cloud
Private cloud has always presented a challenge to enterprise IT on a number of levels, from conceptually figuring out the relationship between private and public cloud, to the nuts and bolts of designing and implementing a private cloud amid resource and staffing limitations.
In the early days of cloud computing, when marketing and other groups were bypassing IT and acquiring SaaS-based business applications on their own, some IT groups simply virtualised a few servers and called that the private cloud.
But the bar for creating a modern private cloud is pretty high. And for a traditional IT department, implementing a private cloud is "incredibly difficult, with a low chance of success," says Warrilow. Here are some of the key challenges:
- The public cloud provides much greater transparency on cost than a private cloud, which enables companies to measure and control expenses
- Private cloud deployments rarely do an effective job of implementing chargeback functionality, which replicates the pay-per-usage model of public clouds
- In order to compete with public clouds, private clouds need to have near instant provisioning, automation and self-service. This is a huge challenge for enterprise IT, which may be dealing with literally hundreds of applications
- Public cloud service providers have scores of engineers innovating on their platform, integrating machine learning and AI into their offerings. This is a resource that enterprise IT can never hope to match
- Private clouds need to offer application developers the same development environment that they're using to build cloud-native apps, the same toolchain, the same automated deployment capabilities. There needs to be a service catalog in the language that developers speak. There need to be available APIs to help developers create as frictionless a process as possible. This is a huge challenge for enterprises, despite the availability of tools such as Red Hat's OpenShift, which is designed to enable containerisation in private cloud environments
An even more critical issue facing enterprises is developing a clear strategy going forward amid contradictory advisories; some analysts are predicting the death of the enterprise data centre, and others are identifying the repatriation of cloud apps back to the private cloud as a significant trend.
For example, Gartner's Dave Cappuccio predicted in 2018 that 80 per cent of enterprises will have shut down their traditional data centre by 2025.
And 80 per cent of 400 decision-makers who took an IDC survey reported that they had migrated either data or apps from the public cloud back to on-prem or private cloud environments. So, which is it?
It's all about the workloads
Forrester's Gardner argues that repatriation is not a broad trend. "It's simply not true," he says. There may be some companies moving a specific application back to the private cloud for performance, regulatory or data gravity reasons, but repatriation is a relatively isolated phenomenon.
The latest Gartner thinking on repatriation is in agreement with Gardner.
"Contrary to market chatter that customers are abandoning the public cloud, consumption continues to grow as organisations leverage new capabilities to drive transformation," reads a 2019 Gartner report from analysts Brandon Medford, Sid Nag and Mike Dorosh.
"Certain workloads with low affinities to public cloud may be repatriated, largely because the migrations were not sufficiently thought through. But few organisations are wholly abandoning the public cloud at any technology layer."
Warrilow says flatly, "Repatriation in net terms is not happening." He adds that there will always be a small number of workloads that go back to the private cloud as part of an organisation's ongoing evaluation of the best landing spot for specific workloads.
The latest numbers from the RightScale 2019 State of the Cloud report from Flexera back that up. In the survey, companies said they plan to spend 24 per cent more on public cloud than the prior year, while spending on private cloud is increasing only eight per cent.
There are many business-critical applications that should always live in a data centre for a variety of regulatory, security, latency or other reasons, and in fact, Forrester's Gardner argues, mainframe usage is actually growing.
Those legacy mainframe apps don't die off, he says, and they should be re-hosted or refactored onto a private cloud platform where functions like provisioning and patch management are automated in order to boost efficiency.
Warrilow adds that there are endless scenarios in which private clouds or edge clouds are appropriate for specific applications – manufacturing plants, refineries, mines, medical facilities, cruise ships, stadiums, shopping centres, etc. Of course, companies can also switch to a colocation facility or hosted private cloud, rather than run their own data centre.
A new generation of private cloud options
Nearly all companies (94 per cent) surveyed by RightScale have adopted some form of public cloud, while around three-fourth (72 per cent) are using private cloud. The percentage of enterprises with a hybrid strategy that combines public and private clouds has increased from 51 per cent in 2018 to 58 per cent in 2019.
Specific to private cloud adoption, RightScale finds VMware vSphere is the leading solution with 50 per cent adoption, which is flat compared to the prior year (although some would argue that simply adopting vSphere doesn't meet the requirement of a modern private cloud).
OpenStack comes in second at 28 per cent, followed by VMware vCloud Director (which is a true private cloud), and Microsoft System Centre.
The fastest growth, however, is coming from the new generation of offerings from AWS and Microsoft. Azure Stack came in at 22 per cent in 2019, up from 17 per cent in 2018. AWS Outposts was announced in late 2018 and was already up to 12 per cent.
Here's a brief synopsis of the new services that promise to bring the benefits of public cloud to the private cloud and create a true hybrid model:
- Microsoft Azure Stack: Allows customers to leverage Azure cloud services from their own data centre. In November, Microsoft announced a preview of Azure Arc, which extends Azure Stack to AWS and Google platforms for customers who want to run distributed environments across on-premises, edge and multi-cloud
- AWS Outposts: A fully managed service that extends AWS infrastructure, services, APIs, and tools to virtually any data centre, co-location space, or on-premises facility. It targets workloads that require low-latency access to on-premises systems, local data processing, or local data storage. AWS has also announced that it plans to deliver a VMware variant of AWS Outposts sometime this year.
- Google Cloud Anthos: Enables companies to modernise applications using containers and Kubernetes on existing internal hardware, build new apps and run them anywhere. Anthos is an open, hybrid and multi-cloud platform
- Oracle Cloud at Customer: Delivers Oracle cloud SaaS, PaaS and IaaS services to enterprise data centres.
For companies that have been struggling to build out a highly automated, highly efficient private cloud, these new options enable private clouds to reach parity with public clouds in distributed cloud scenarios where each workload runs on the optimal platform.