Hybrid Cloud is a myth.

It’s fiction, IT folklore, snake oil.

This is the kind of “hybrid cloud” people seem to be dreaming about:

A kind of hybrid cloud where their (any) private cloud is a continuous fabric extending into (any) public cloud(s) managed seamlessly through single management/control plane.

In the dream, they achieve agility, scale, cost savings, and no lock-in as their workloads seamlessly and automatically fly back and forth seeking out the most appropriate cloud based on demand, cost, performance, security etc.

I love this dream too, folks, but I hate to break it to you:

Such a beast doesn’t exist — it is a Unicorn.

To understand why, let’s examine what is really going on in the IT landscape today

This is more or less what it looks like:

One or more public clouds

Used for most cloud-native, new-age applications (mostly “cattle” in pets vs cattle), demos, PoC’s, training, and a healthy mix of production-ready long-running workloads and temporary workloads.

AWS dominates this market with Microsoft (Azure) and Google Cloud trailing behind

On-premises Enterprise IT environments

Used mostly for legacy applications (“pets”in pets vs cattle). These environments consist of heterogeneous and hairy mess of bare metal, VMWare, Microsoft, KVM/Linux virtualization. Furthermore, they are deeply integrated into the enterprise database, security, networking, compliance/identity management stack. “New” applications in this space will look a lot like legacy applications (same people, process, and culture — cloud will not change that!).

These environments and workloads represents 90% of the market today but they are also the most architecturally incompatible with the public cloud

Some small private cloud environments

Used mostly for development and testing workloads, content delivery, or centralized storage. These environments are most likely built on VMWare or OpenStack and will mostly survive in niche regulated industries like health care, Biotech, Defense, Govt etc. However, for the broader enterprise and larger scale, it is just too much hassle and expensive to build and operate on-premises private clouds.

My previous blog post Private Cloud is Dead presents a vendor/market (supply side) perspective on why this will remain a small stagnant market

A few large-scale private cloud environments

Many telecommunication giants will be forced to build to address the 5G challenge and to not get commoditized as “dumb pipe” vendors. They can’t use public cloud because this is their core business and they can’t use traditional hardware networking or traditional virtualization because it is not flexible, scalable, or cost-effective enough. For similar reasons, we may find few large banks and retailers build their own private large cloud environments. They have the wherewithal, deep pockets, and urgent business pressures to make it happen.

A few observations about these environments:

• The architecture, networking, security, performance, availability and reliability characteristics of these environments tend to be radically different from each other.
• One or more of the above environments may be in use simultaneously within the same company, usually in isolation.
• Workloads that run in an environment typically tend to stay in that environment.

You may want to integrate on-premises environments with the public ones for good reasons, but getting these environments to work with each other requires significant development, testing, system integration , and on going support and maintenance.

And if you need an army of people or consultants to integrate, maintain and support, that ain’t “Cloud”, my friends.

But, wait a minute — you might be tempted to say — I can automate all of that integration and bring in a whole bunch of tools to make it work. Sure you can move VM’s from one environment to another using migration tools and you can use a single cloud management platform (CMP) to control/manage resources in all the different environments and when container technologies mature, you could probably achieve 100% workload portability.

Yes, many of these tools exist today. True. However, all that integration work/effort with all that new tech and tools is extremely complex and takes a lot of new skills to pull it off (remember process, people, culture?). This was the reason why private cloud failed in the first place and now you want to take it a whole new level by adding more clouds to the mix? Really?

Wait, there’s more. There’s a much, MUCH bigger challenge:

The hardest thing to do is to ensure SLA’s and integrity of your hybrid workload when you are dealing with a series of completely different and incompatible environments (owned by different entities) over which you have no control.

And here’s the kicker:

The combined availability of two separate clouds used together is always lower than the availability of the individual cloud. Say for example your private cloud offers 99.99% availability (4 nines) and your public cloud offers 99% (2 nines) availability. The combined availability is the product of the two i.e 98.99% which is less than both your private and public cloud. Add another cloud with 90% (1 nine), your availability drops further to 89%! Your downtime per year went from 52 minutes to 40 days, OUCH!

Is there a type of Hybrid Cloud that can work?

Sure there is. It is one where the on-premises and public environments have been pre-integrated/pre-packaged or alternatively architecturally compatible and managed by a single provider who can provide industry SLAs.

Architectural compatibility comes from hybrid offerings such as Microsoft Azure Stack (works seamlessly with Azure public cloud) and Oracle Cloud Machines (exact replica of Oracle public cloud).

While pre-integrated and packaged hybrid solutions come from such hybrid offerings such VMWare on AWS and Salesforce on AWS.

In case you haven’t noticed, these are all proprietary platforms. If you are screaming “LOCK-IN”, welcome to the brave new world of our cloud future. What matters is generating business outcomes with strategic agility and these platform promise just that.

It’s just what way we roll in our beloved industry. Enjoy the ride.