20 Years Into Cloud And Still Hardly Anyone Fully Understands It

If there has been one overriding theme to IT delivery in the last 20 years, surely it's been "it's the cloud, stupid".

Yet, most IT people I meet understand only a few of the "cloud's" basic capabilities but hardly anyone really understands the big picture, the fundamentally transformational nature of "cloud computing" and where it's going. Some of the basic capabilities that "cloud" offers that most people understand are:

• quick and inexpensive basic resource provisioning

• the ability to fine-tune the resources horizontally (in time) and vertically (size) to match the actual usage needs

• multitude of available business services in the form of computing components by the cloud provider and in the form of SaaS by vendors

• predefined architecture patterns matched by preconfigured templates to build new services and applications

• centralized provisioning, security, logging, monitoring, support

• all of the above is available almost at any scale for hundreds, thousands, or even tens of thousands of resources

All the points listed, and #1 above all, are the major selling points of the "cloud" but is that all? Today, infrastructure procurement is just the most basic and simple service that a cloud provider offers. Initially way-back-when 20 years ago when the cloud was starting this may have been the main purpose of the cloud companies. But today, 20+ years later, it's just a basic precondition to cloud computing.

What seemed like a revolutionary idea, a fast and cheap infrastructure provisioning, may not even be a significant real-world advantage today. I remember an article on AWS a few years back (I cannot find it now) that compared the operational costs of an on-premise data centre with AWS. Since it was published by AWS it obviously concluded that it was cheaper to run your IT in AWS but the actual numbers were quite close. A fully optimized AWS cloud would save a company at best 20-25% of the costs over 3 years, if I remember correctly. While 25% saving is not insignificant for any organization when you add to it the cost of migration, the risk, and the new skill set that needs to be acquired, it's not unreasonable to conclude that the whole cloud thing is just not worth it. And that's exactly what many organizations in Canada have concluded.

But the above scenario is the ideal one where everything is either in the cloud or on-prem. In the real world, this is rarely the case. In the real world, companies run hybrid environments meaning they use on-prem and cloud environments with various percentages split between the two.

Yet, hybrid is the worst of both worlds, it's more complex and more expensive than either on-prem or cloud alone because it combines both. I have seen strategy papers implying that hybrid is the target architecture for the specific organization. When I point out that it doesn't make sense due to higher costs the strategists tell me that they cannot foresee their whole IT being run from the cloud for various reasons. Then I suppose one must assume that going to "part-cloud" is because everybody is doing it, a classic business strategy "me too". It's not just a lack of imagination but the fact that these high-level managers don't understand the capabilities of the cloud fully, they understand the bullet points but not the transformational nature of the cloud and where it's going.

In basic architectural terms, a hybrid cloud is the transitional architecture and a full cloud is the target architecture even if it takes a thousand years. It is the only logical architecture.

So what is that "transformational" nature I keep talking about? It's not another bullet point I missed above, it's more than that. All the bullet points above could be classified as attributes of the standard Cloud service types or levels:

1. IAAS - infrastructure as a service

2. PAAS - platform as a service

3. SAAS - software as a service

But there is a type # 4 that hasn't been named yet and that hardly anybody understands except for the fineschmeckers of cloud computing. It's this level that makes the cloud something different from "Rackspace", one of the earliest cloud providers of level 1 and 2 services (Rackspace was no AWS).

The above levels are in a way levels of abstraction (progression of abstraction) or we can call them softwarization of computing. The highest level (though lowest by the numbering scheme as #3) is SaaS which is simply a software service without anything else as far as the user or client is concerned. SaaS means a single software package but what if all the software is SaaS, in other words, if an organization is using nothing but SaaS and the SaaS packages are connected by SaaS services we have achieved something quite different from an on-prem or cloud environment on their own or combined.

We have achieved infrastructural "singularity" (everything is one thing, the link in jest), everything is a service and there is no hardware/software to be managed anywhere in our domain. That's level #4 and let's call it ComputingAsAService or CaaS.

4. CaaS - computing as a service

This is not only possible now it's the actual logical culmination of the cloud revolution. It's the necessary outcome in the future.

Some will say that it's impossible to have a pure CaaS (all SaaS) environment because one needs custom software which is by definition not SaaS. I think they are wrong, CaaS will be a reality though not soon. Eventually, an organization will not purchase software packages but a single unified computing service that will include all the SaaS an organization would need such as desktops, CRM, accounting, HR, inventory, internet storefront etc. as part of a service offered by the vendor.

But even if an organization must develop its own software to plug the gap between SaaS offerings from the vendor, the organization should develop this as an internal SaaS service. The way to do this is by using cloud-native architecture called SAM (serverless application model). This architecture and the software built upon it uses only services and hence is itself a service. This is the ultimate custom-built software architecture which is cleaner, easier to build and way cheaper than the traditional server-hosted software.

But this alone though significant is not transformational, that is, the fact that in, more pedestrian terms, infrastructure is turned into code (IaC on the provider side) is advantageous but is not a business revolution. It needs another ingredient to make it a game changer. And that ingredient is the cost.

From the cost perspective, the only viable cloud is a full cloud with some regulatory exceptions. The cloud costing model is such that the more computing the client does the bigger the discount, hence, the price goes down when calculated per computing unit. This is very different from the on-prem environment where buying additional resources increases the costs linearly. And with SAM as an internal SaaS, the provisioning costs are so dramatically lower that it truly is revolutionary. For example, when I took a course for AWS architect certification, the instructor said that their education courses were all SAM-based and that their cost for serving the lessons was a few hundred dollars per month compared to over $100K per month when using cloud virtual machines. Hello, anybody home? How many IT managers understand this figure? My guess is < 1%.

I used the term "softwarization" for this process where the end result is a universal computing service but there is another term that is used in capital markets and that may be more familiar to readers. It's called commoditization of the sector, specifically computing (IT technology). It means turning a highly complex high-cost production of a product or a service into a simple cheap commodity that is delivered in bulk (at scale) for a very low price based on the volume. That's what computing is ultimately going to become where price is based on volume and not a specific individual component price that is scaled linearly.

Hence "cloud" computing will ultimately become a commodity that will be sold and bought in bulk cheap and plentiful. A client will buy an accounting service available online, not a data center that houses a network with a connected computer with accounting software installed and supported by multiple people in different roles.

And that's the message for the CIOs who still haven't grasped and haven't accounted for this in their strategies and roadmaps. Simply put, there is no commodity to be had in hybrid cloud and even less so on-prem.