Is virtualization different than cloud?

Disclaimer: The below is My Personal Opinion (MPO) and likewise may not be shared between me and my employer. This in no way is to be considered the opinion of my employer or any subsidiary of the larger organization.

“Which came first, the Chicken or the Egg?”

It’s an age-old question, that depending on which side of the fence you’re on is to which you’ll attest to. For Virtualization and Cloud there is no doubt which came first, it was virtualization, and what Cloud is (MPO) is the evolution of virtualization. Nothing more, and if someone wishes to explain differently I would love to hear about it!

That said, let’s start at the beginning of this evolution: Virtualization, is a software layer (Abstraction Layer) which separates the underlying physical infrastructure to create a standard, stable set of leverageable resources. In doing this the underlying hardware doesn’t need to consist of the identical hardware across multiple Virtual Infrastructure (VI) hosts to support the Operating System Instance (OSI) or Virtual Machine (VM) which will run atop of the VI. Why this is important is due to the ability to procure commodity IT hardware as forecasted by the business all the while providing a stable, scalable, Highly Available (H/A) platform on which to host the business’s applications/services. Resiliency or H/A capabilities are provided by the VI by performing resource based consumption balancing (Optimization), Predictive Failure Analysis (PFA) along with other fault-tolerant and recovery capabilities dramatically reducing degraded or loss of services which can have both financial and reputational impact to the business.

Beyond the hardware abstraction, virtualization also provides the ability to concurrently run multiple OSI’s on a VI host while providing security with binary isolation to gain levels of efficiency in reducing IT spend while increasing utilization and agility of the acquired IT assets (Running multiple OSI’s on a single physical VI host is also known as workload consolidation.).

This software layer also known as the Hypervisor provides an allocatable set of hardware resources (Including but not limited to: CPU, RAM and Network.). The underlying physical resources are by default ‘not’ dedicated and as such offer the capability to be allocated to more than one OSI concurrently (known as leveraging.). By leveraging these physical resources, businesses can accomplish much higher percentages of utilization, moving from an estimated industry average of 10-20% to 55-60% or potentially higher. Depending on how aggressive (Risk Tolerant) the business wishes to be.

But isn’t there a missing main ingredient to the defined resources above (eg: CPU, RAM and Network)? Well aren’t you astute! You see I made the statement that the above resources were ‘not’ dedicated upon allocation. That would not be true for storage where just the opposite is true. By default, it is dedicated to the OSI to which it is provisioned and only upon explicitly setting the storage to shared would it be so. Now, before this turns into a rat hole the overall pool of storage (SAN, DAS, Hyper-Converged, etc.) is ‘not’ dedicated, nor assigned to the individual OSI, it is only a logical chunk/slice that is dedicated. The pool of storage is a leveraged resource as are all the above.

I would say the above is enough detail to gain a level of understanding for Virtualization, now let’s discuss the evolutionary leap from virtualization to cloud.

What is the meaning of Cloud? Should be the first question needing to be answered…

Cloud (MPO) is the commodification of compute resources. That is the shortest possible answer. To expand on that Cloud is an enabler for businesses to move from a CAPEX to an OPEX model by eliminating the business from having to obtain and maintain hard IT assets and the facilities to host them. Now there are going to be a number of people that this raises a hair or 2 with who will say, but that’s not all it does! And I will say, that makes it a differentiator, yes it is!

Let’s drill in one level to Cloud. There are 3 basic types of Cloud namely: Hybrid, Public and Private. Below is a quick description of each:

Hybrid Cloud: The easiest to explain, take a portion of Public and a portion of Private shake gently (Integrate) and you have hybrid (Pun intended!).

Public Cloud: A fully commoditized virtualization platform, where all resources are provided with customization to fully integrate into each business’s unique requirements. Leveraging a utility based model where the business is billed on a consumption basis. There is no cost beyond the consumption and this consumption could be on a per minute, hour, day or larger quantitative unit of measure.

Private Cloud: Same as above with a caveat as the Business is potentially still running in a CAPEX model:

• Business owns/leases the IT Assets.
• Business owns/leases the facilities to host the above assets.
• Business needs to retain the resources and services to maintain the above IT assets and potentially the facilities.

Now I am sure you are thinking, that was an unusual way of presenting that (sort of a bottoms-up approach.), and to that I would say yes, yes it is. But! There was a reason for it. Private Cloud doesn’t just fall into the above description. If it did my expanded definition of cloud would be incorrect. There are service providers (SP) which take Private Cloud and move it into a utilitarian model for businesses so they have none of the responsibilities for owning the IT assets listed above, nor the facilities and potentially no human capital (resources) to maintain, as all of this is provided by the SP. The difference with this compared to public cloud would be that these IT assets are dedicated and usually there are minimums contracted which provide the SP the ability to recover their capital investment in providing the IT assets. Secondly in the hosted private cloud you have potential visibility to the IT assets (Depending on the selected VI platform.), whereas in public these are all leveraged and need to be considered “Black Box”. Where this becomes a highly viable solution would be where data sovereignty, regulation or other governing mandates reduce the ability to leverage public cloud. One other area where this is a viable solution is where things such as high utilization come into play as with most public cloud providers Network and I/O have incurred costs and at some point, there will be a threshold where the cost of public surpasses private.

OK, back to my opinion above commenting to Cloud being the commodification of compute resources and moving businesses from a CAPEX to OPEX model is the only differentiator…

I’m would expect responses like:

“But Cloud makes for a highly agile platform for applications/services to be hosted.”

My response: Yes, just as the virtualization could, that is presently in your business’s Data Centers.

Further, these Data Centers could be privately owned and operated, or hosted in a Co-Lo. The difference being that in the Cloud those Data Centers and the incurred cost for providing the Roof, Power, Environmental Controls (Humidity, Temperature, etc.) and Management/Maintenance are not a burden carried in whole by the business. Nor would they for the Servers, SAN’s, Network Switches, Routers, Firewalls… the list goes on and on for the devices that all require the acquisition and ongoing maintenance to retain a stable infrastructure.

“But Cloud enables quicker time to provision and scale (Up and Down as needed.).”

My response: Yes, just as the virtualization could, that is presently in your business’s Data Centers.  (I’m see a repeating pattern here!)

The same level of automation and integration available in the Cloud is available with your business’s virtualization, only the majority of business took the quick win with consolidation and just continued to move forward with the development of their business. Neglecting the potential efficiencies obtainable in taking a holistic look at their infrastructure and automating it.

Another aspect to this is Business controls. For example, developers (Not pointing anyone out!) love agility: quick provisioning, rapid coding (sprints), functional testing and move into service (deployments). While this sounds great, working in a bubble and not having the right controls to go beyond functional testing, can lead to catastrophic damage to the business, and these controls come at a cost of time and efficiency. Now this is not to say “Suck it up and deal with it!” not at all, as there needs to be a holistic review of what policies and procedures are implemented to make sure it aligns with your business’s needs, not just a copy and paste out of the ITIL standards. ITIL it is not a rule! It’s a set of guidelines, to which the individual business needs to use to define their policies and processes and these will be unique to each business.

“What about PaaS, SaaS?”

My response: Yes, just as the virtualization could, that is presently in your business’s Data Centers.  (I’m definitely see a repeating pattern here!)

PaaS and SaaS are still based off the underlying Virtualization. For example lets look at a PaaS instance of MS SQL, Let request it from Azure. What did we get, we got in a matter of minutes an instance from an existing AlwayOn SQL Cluster which is either provisioned on a physical or a virtual construct (Being that Microsoft does not make this information publicly available I can clearly state it one of those two!) In either case, this functionality can be provided by the on-premise virtualization. Let’s look at a SaaS application formed from a combination of Containers (Docker and Kubernetes) again all hostable from the business’s existing Virtualization.

What I would be surprised to hear is:

“But Public Cloud provides 99.5% Uptime, which my business’s VI cannot meet”

My response: Yikes, That’s terrible! Because 99.5% uptime isn’t all that good! In reality that equates to 3.6 hours of downtime each month. And that is an infrastructure based SLA. It does not encompass patching of the O/S for IaaS hosted applications/services or your application code hosted in an I/S/PaaS solution, all of which you need to subtract from that 99.5 % uptime SLA.

Public Cloud takes away controls you presently have. Am I saying that is a bad thing? No, but just like anyone looking to truly adopt a Cloud First strategy your mindset ‘must’ change! Application Owners no longer have the capability to point the finger at the CIO or his Infrastructure team when it comes to an application outage, as they no longer control it! It’s a black box to all but its provider. Application resiliency is no longer dependent on the underlying infrastructure. It needs to be incorporated into the way the application is architected. Leveraging Fan-out or application pooling eliminating all potential Single Points of Failure (SPoF). Establishing Affinity rule ensuring the application instances are distributed across multiple hosts or locations mitigating all possible disaster possibilities. Using available services such as Load Balancing to manage traffic Local across multiple VI Hosts and/or Regionally/Globally. And if, your going to apply this different mindset to the cloud wouldn’t it also apply in your business’s current virtualization?

So, that was a long-winded way to conclude that virtualization isn’t different than cloud, it is in fact the foundational stepping stone for cloud, without it cloud would not exist as we know it today.

Now that I'm confident I've ruffled some feathers, what I will add is where Public Cloud has an advantage over Private is that they have the capability to provide the agile, efficient, integrated I/P/SaaS platform today. Whereas Private Clouds do not. This I can say is the case for likely 90% of all Private Cloud implementations with a high level of confidence. And that the other 10% are the technology evangelist, the forward looking, thinking businesses that never stop thinking of ways to improve. Oh! also in that 10% are some of the SP’s that provide hosted private clouds which in some cases may place them ahead of Public Cloud as they should have a more robust service catalog as they have worked with numerous customers and if correctly approached they have used a development model where the code written for one business’s needs is written in a way to commoditize it for reuse by all their customers.

Thank you for taking the time to read my post. If you liked it let me know and if you have a difference of opinion, likewise please share it.

After 20 plus years in IT I know there is plenty more to learn and am always open to listening to others opinions or criticism (Positive or Not Positive) which will only help me in the long run.