The State of Hybrid Cloud in 2017
Hybrid. As far as I can tell, this means the combination of two things that are not the same to serve the same purpose. Like hybrid cars — half electric, half fuel. Or like hybrid energy — half nuclear, half solar. And of course hybrid, as in “Hybrid Cloud”. So this cloud-related usage of “hybrid” has been around for years and is one of the most misunderstood and most weirdly interpreted terms.
In this blog post I want to discuss the — as far as I have witnessed — history of the “hybrid cloud” and my personal view on the outlook for the next 2–3 years.
Hybrid Hardware Hosting
In 2014, when I joined the cloud world through a change of teams at CSC Germany, the journey beings. At CSC, we had our own public and private cloud business running in parallel. Both services looked quite promising. It was VMware based hosting environments. Of course, there was already this AWS around, but we competed mainly with our traditional competitors competing (HP, IBM, you name it). So that was that.
The whole fuzz about “hybrid” began to emerge out of the challenges that we faced (on paper) in larger outsourcing deals, where we tried using both public and private cloud services to create cheaper and more flexible customer solutions. What that meant in reality was: dig deep into network connections and storage clustering. The focus was on layer-2 stretching and disaster recovery replication. We were hybrid on the hardware and hypervisor level, nothing else. It seemed natural.
However, we soon found out that a) this is not really a value add to the end user and b) it is a very expensive and complicated setup to build. The key challenges there were about having a (very) homogenous environment on both ends — even today this is nearly impossible (see also: VMware Cloud on AWS), as the public cloud service provider always updates faster than the on-prem hosting operator.
Hybrid = private cloud + AWS
Around 2015 we saw public cloud vendors, like AWS, overtaking the public cloud market entirely, forcing us to sunset our own service. How sad that was. But also, how promising the future looked — now we could leverage the amazing speed and agility of AWS and combine it with our private cloud VMware based portfolio. The pieces of the puzzle seemed to finally fall into place. Now, a new challenge came: how do you properly govern and orchestrate these two platforms in a “unified” manner. Countless software companies were created and sold to larger SI’s to enable this “game” and help solving paper-tiger problems once again.
By the time I left CSC, there were almost zero customers that ever really cared about managing or orchestrating their AWS and VMware farms alongside or through some single pane of glass. Because it wouldn’t work properly as soon as you viewed AWS as more than just a hosting farm. Period. This fact remains unchanged until this day.
Hybrid as in: no private cloud at all!
In 2016 then, when I joined Nordcloud, the term “hybrid” was still present — but used for a different thing all of a sudden. Being pretty much public-cloud only, the Nordcloud people did not use “hybrid” in reference to public-private mixery, but to describe “AWS + Google + Azure” setups. These are of course also hybrid, if you ignore the existence of anything on-prem. Sigh. So there I was, yet another definition to work with.
The good thing today is that while any private cloud on this planet does so far not really work well with any of the three leading hyperscale platforms — if you cross out private, you get something quite homogenous there. We very regularly see customers adopt both AWS and Azure and — use case driven — deploy workloads through a unified governance and best practice model into both clouds. The remaining challenge of course is lock-in with PaaS and SaaS features of both providers. But that is a far more advanced challenge.
What has changed in comparison to 2014 is that “hybrid” no longer refers to “physical and logical connections” but rather “using different platforms the same way”.
The Future of Hybrid Cloud
Regardless of the past, there is but one certain thing: more interpretation of the term is to come as soon as products like Azure Stack really hit the market. Another big impact will certainly come from stuff like Kubernetes, offering a cloud-to-onprem cluster to overlay / stretch across infrastructure platforms.
No matter what you build and design for, you will never be able to overcome the natural limitations and implications around networking and storage, as well as consistency and latency. Some use cases and scenarios are just never going to fly. But as long as there as a technical and user demand to make hybrid happen for a real reason, there will be a budget and a solution.
P.S.: with the rise of big data and data lake scenarios in the enterprise (as in actually happening right now), we are going to witness the relevance of Hybrid Cloud growing again to where it once was, as most of the data can’ be put into AWS or Azure for legal and compliance reasons → Hybrid Cloud Architecture needed.
