If there’s one safe bet in the IT industry, it’s that technology adoption will be a messy process that rarely follows a quick, neat storyline. We see broad trends over time: standardisation, open source, distributed computing, mobility, virtualisation, and most recently containers and associated cloud-native tech. But any claims that some particular approach has “won” or will win must inevitably be tempered by a recognition that technology adoption is unevenly distributed and needs to serve a multitude of requirements and masters.
We’re seeing this playing out with cloud computing. I’m almost tempted to call 2018 the Year of the Messy Cloud or the Year When Clouds Embraced their Differences.
This isn’t really a new trend but consider where we came from. Cloud computing was first seen as an analogue to the electric grid, a commoditised utility generated by centralised producers and generically consumed. Public cloud providers were never as standardised and interconnected as the grid. Nonetheless, various open source projects sought to abstract the differences between cloud interfaces for services like compute and storage. Others envisioned service brokers that would move workloads around between clouds to optimise pricing on even a minute to minute basis.
These ideas were born of a period after the dot-com bubble popped (the mid-2000s or so). IT was seen in many circles as primarily a commodity; meaningful differentiation was the domain of a small number of specialised providers with most of the complexity hidden from the users of the services.
But, for the most part, things haven’t played out along such a constrained and monolithic path. Which really shouldn’t be surprising. They never have in the history of the computer industry even if some technologies dominate for some tasks for some period.
As a result, many of us have talked about hybrid clouds almost from the beginning. The fundamental idea of hybrid clouds is that organisations will mostly have a variety of infrastructure footprints including physical, traditional virtual, private cloud, and public clouds. At the same time, hybrid suggests a degree of workload portability, unified management, and even workloads that span across multiple clouds (shades of client/server!).
A simpler, broader concept
One of the changes that we’re seeing today is that hybrid cloud is not giving way to but it is being married to the concept of multicloud. The multicloud concept is simpler but broader; multiple clouds either on-prem or with a public cloud provider.
In a sense, this is a further recognition of IT complexity. Hybrid concepts like portability, freedom from lock-in, and common management remain extremely important just as standardisation within datacentres has been a driving force for technology selection for many years. Market researchers IDC have recently published a study that found accelerating cloud repatriation, i.e. workloads moving from public clouds back to on-prem. This is strong evidence that, in addition to workloads moving to public clouds, they’re also moving from those clouds. The obvious corollary is that you want to make technology decisions that facilitate such moves. These include using standard operating system environments like Linux, containers and container platforms, avoiding cloud-specific features where they don’t add differentiating value, and generally avoiding placing workloads and data where they’re difficult or expensive to get out again. (For example, egress charges for large data sets in clouds can be significant.)
At the same time, modern multicloud recognises that the world is complex and messy. Cloud providers provide unique features in areas like machine learning that may justify tying yourself to that cloud provider for an application that may not be very long-lived anyway. Many legacy applications will remain as on-prem monoliths for the foreseeable future because there’s no business value in rewriting them. Specialised online platforms for office productivity, communications, analytics, customer relation management, and many other purposes may offer benefits relative setting up and administering your own more portable application.
These systems all need to talk to each other to various degrees. But there’s a distinction between clouds that are fundamentally hybrid and connected, if not in the most purist sense--“seamless interoperability” has essentially always been an aspiration rather than a reality--and the broader multicloud landscape.
In any case, whatever the particulars, hybrid clouds and multiclouds have emerged as the dominant pattern whereas going all-in on a single cloud is rare. Market research firm Gartner tells us that: "The landscape of cloud adoption is one of hybrid clouds and multiclouds. By 2020, 75 per cent of organisations will have deployed a multicloud or hybrid cloud model."
It was once the case that IT departments needed to piece together a wide range of hardware and software components to make their datacentres tick. They still need to do so, at least to a certain degree, especially in the case of software. On-prem clouds and container platforms running on software-defined infrastructure are increasingly prevalent.
But IT also needs to approach its role as a broker of services which chooses from among the many clouds on offer and makes the hard decisions about cost, portability, risk, governance, security, capability, business benefit, and more.
Gordon Haff, Open Source Evangelist, Red Hat
Image Credit: Everything Possible / Shutterstock