Recent signs of detente notwithstanding, cloud service providers and traditional enterprise IT equipment vendors regularly engage in a tug-of-war of press releases and case studies touting the latest major company to either: (a) abandon their data centers for the public cloud, or (b) recoil from shared cloud services in favor of shiny new private cloud infrastructure.
For every GE going all-in with AWS, it seems there’s a Dropbox leaving (at least in part) for the greener pastures of a private, self-managed environment. The latter phenomenon has a memorable, if not controversial new label, cloud repatriation, as if like Dolly in the musical, the workloads are going back home where they belong.
It should come as no surprise that incumbent IT suppliers are big fans of cloud repatriation. Dell EMC has promoted the concept and it was a hot topic at the recent VMworld event, including at the annual research briefing by IDC where analysts shared some survey data and their conclusions about a looming surge in enterprise private clouds.
Meanwhile, hard financial data shows that public cloud usage is booming, so there's undoubtedly something more to the story. (sic)
Workloads moving to the cloud, the question is which one
The current case for public cloud abandonment rests on a January 2018 IDC survey finding that this year, 81 percent of 400 "IT decision makers" will migrate applications or data "that were primarily part of a public cloud environment to a private cloud or on-premises environment." The number jumps to 85 percent when considering plans for 2019. The nuance is that the destination for these workloads is anything but uniform.
For example, IDC reports that 20 percent will return to a traditional on-premises, non-cloud environment, while another 43 percent are destined for a retooled on-premises private cloud. That leaves 37 percent going elsewhere, which IDC deduces will be hosted private clouds, a fraction it expects will increase to 47 percent in 2019.
IDC further estimates that managed service providers will spend an incremental $37 billion between 2019 and 2022 on hardware for new hosted cloud environments, which will generate at least $40 billion in subscription revenue plus untold more in multi-cloud management services.
The implicit assumption is that such repatriated clouds will resemble a cross between public cloud services and traditional data centers, namely large, consolidated equipment farms in colocation facilities that runs one or more enterprise cloud or container stacks like VMware, Azure Stack, OpenShift or Cloud Foundry.
As I discussed earlier, such an infrastructure-agnostic meta-cloud is the vision VMware touted at its recent event. However, as I also pointed out in my last column, there are legitimate reasons of performance and cost that many of these workloads will end up moving to non-traditional edge locations near the sources of application data and/or users.
The likely evolutionary path of edge cloud services is unclear, however as my column detailed, the system Vapor IO is building resembles a massively distributed colocation service using modular data centers the size of shipping containers positioned at cellular base stations or other strategic network locations.
Should an organization move data-intensive applications from a conventional public cloud to such a managed edge environment, it fits the accepted definition of cloud repatriation, even though such environments bear no resemblance to traditional enterprise data centers.
Squaring repatriation talk with public cloud business results
A significant problem with surveys such as IDC’s is that they measure IT intentions or aspirations, not results and as such are incapable of verification. However, as the saying goes, the road to
troubles nirvana is paved with good intentions.
In contrast to what people say they will do is financial data reflecting real-world IT activity. Since AWS and Microsoft have begun breaking out cloud revenues, we have good evidence of how aggressively IT departments continue to spend on public cloud services.
For example, quarterly revenue at AWS has increased six-fold over the past four years, a CAGR of greater than 50 percent, to an annual rate of more than $20 billion. Microsoft doesn’t break out Azure revenue, however, it did report an 89 percent increase in its Q4 2018 report, with one financial analyst estimating that Azure brought in more than $2 billion during the quarter for an annual run rate rapidly approaching $10 billion.
It is conceivable those numbers can be explained by a general increase in computing need but that does not show up in other analyst numbers.
Angela Lambert at TBR Research pivots off of the cloud revenue data, along with its estimates of overall growth in the IaaS and PaaS markets, to dispute the notion that organizations will soon be abandoning public clouds. While admitting that cloud repatriation is happening, it’s “not real enough to change the prevail line cloud trajectory” and is “the exception, not the rule.” TBR’s survey finds that 37 percent of respondents plan to increase their use of public cloud over the next three years, a higher proportion than those increasing use of hosted or self-managed private clouds.
She makes a good point about the repatriation talk neatly fitting into a narrative that IT equipment vendors would like to promote in the face of cloud services eating into their fat margins, writing,
If your job is in any way related to selling products or services for enterprise data centers, ‘cloud repatriations’ sounds like a promising concept. Amazon Web Services (AWS) and Microsoft have been eating the lunch of a whole bunch of IT companies, and those IT companies would like that lunch back, thank you very much. But is the exodus of customers from public cloud really happening? Well, I have some good news and some bad news.
Lambert concludes that cloud repatriation “is not a market changer,” but that companies also aren’t abandoning on-premise data centers. Another TBR analyst adds that to the degree repatriation happens, it won’t be a one-way street, noting that while some enterprise workloads that have moved to public cloud will return to self-management:
...there will also be services deployed on premises that could eventually be moved to a cloud environment as customer needs and costs change.
451 Research finds that when it asks its enterprise IT survey respondents about the reasons for moving workloads back from the public cloud, the reasons are identical to those cited for moving to the public cloud in the first place. 451 likewise concludes that the repatriation story is more nuanced than one of public cloud abandonment, but rather an evolution of enterprises to a multi-cloud/hybrid cloud strategy.
My take: Application and infrastructure evolution, not migration
Cloud repatriation is a semantic deception since it implies a return to some former state of being. Enterprise cloud usage is more like an evolutionary process in which form and function change over time.
Lost amidst the repatriation talking points is IDC's guidance on the topic suggesting that next-generation private clouds don't mean "everybody having their own cloud," but entail delivering cloud services on dedicated systems. Thus, from a server standpoint, they are more like AWS Reserved Instances and Dedicated Hosts than a traditional enterprise server farm.
Likewise, IDC contends that unlike yesterday's on-premise infrastructures, the best new private clouds will provide rapid, flexible deployment and expansion. I would add that "flexible" doesn't just mean the ability to quickly adjust usage and capacity based on workload demands, but to support deployments in non-traditional, decentralized locations, i.e. an edge cloud.
These attributes expose a notable contradiction with the conventional notion of a private cloud since flexible, scalable capacity and consumption is the hallmark of shared, pooled infrastructure, not dedicated resources. Furthermore, modern applications are no longer monolithic entities running on dedicated systems, but mash-ups of higher-level application and data services, SaaS products and custom code running in application containers or serverless functions. Such composite workloads can't be provisioned solely on private, dedicated infrastructure.
In essence, I think that new, "repatriated" private clouds will often end up being owned and managed (at least at the infrastructure level) by a third party, whether that's an MSP, one of the existing mega cloud vendors or an emergent edge cloud providers like Vapor IO, and built from reserved or dedicated services that are allocated from larger shared resource pools.
As enterprises develop infrastructure strategies for future applications, they shouldn't let the repatriation badge for private cloud muddle their thinking and look past old concepts of dedicated, private infrastructure to a new mix of managed cloud, platform and software services that can be consumed both centrally and locally.