The Realities of Cloud Computing Deployment Models


In the past few months, I’ve attended several technology conferences, presentations and MeetUps, and it seems that everyone is using “The Cloud” to deliver technology services to their businesses. But with so many definitions for the cloud, what are businesses really getting?

What is the Cloud?

Last week I googled “What is The Cloud?” Most of the answers returned were geared toward end-users and consumers, indicating that it’s the place to keep all of your personal stuff (e.g. photos, videos, docs, etc.). So I turned to the National Institute of Standards and Technology (NIST) for its definition of cloud computing:

“a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.”

This model includes five essential characteristics (on-demand self-service, broad network access, resource pooling, rapid elasticity and measured service) and four deployment models (public, private, hybrid and community).

This got me thinking: How do the deployment models compare in terms of satisfying the essential characteristics? Since my research in this area came up short, I thought I’d share some of my own thoughts on the topic.

A Comparison of Private and Public Clouds by Essential Characteristics


On-Demand Self-Service

The tendency with private clouds is to restrict management to the IT department. Relinquishing control to a finite set of resources (e.g. servers, storage, network bandwidth) tends to be the biggest barrier to embracing on-demand self-service in private clouds. However, many businesses have embraced the DevOps culture, breaking down this barrier, at least for their software development teams. Public cloud providers make it simple for anyone with a credit card to manage infrastructure resources through GUIs and CLIs. However, most businesses that utilize public clouds are likely to restrict resource management to the IT department.

Does your company allow HR representatives, the finance and accounting teams, or your software developers to manage their own infrastructure needs?

Broad Network Access

Private and community clouds are usually restricted to the company’s WAN/LAN. While BYOD and company VPNs are breaking down these restrictions, private clouds will likely never match the any device, anywhere on the Internet paradigm. Public clouds are available from any device, anywhere on the Internet.

Resource Pooling

In most private cloud implementations, resource pooling is limited by the architectural designs that maintain a separation of production, staging, development, business systems, data warehouses, etc.

Given this limitation, I wonder if businesses are really building private clouds or are they merely referring to their on-prem and hosted data centers as private clouds?

Public cloud providers have solved a problem that businesses have struggled with for a very long time: how to get 100% utilization out of the IT infrastructure. While it may appear as though the public cloud consists of an infinite amount of resources, the reality is that optimizing those resources so that a production application that runs at peak CPU utilization during the night runs on the same physical hardware as a development application that runs at peak CPU utilization during the day. Or my CPU intensive app runs side-by-side with your memory-intensive app. In any scenario, the provider guarantees a specific SLA, which alleviates the consumer from constantly worrying about having enough capacity, assuming the SLA satisfies business requirements.

Rapid Elasticity

When considering the elasticity of a private cloud, the first thing that comes to mind is: define rapid. Elasticity is going to be severely limited by business processes, because private cloud investments are capital expenditures, which usually require department and finance approvals and appropriate planning during the budget cycle. In addition, the physical aspect of the private cloud introduces lead times for resource acquisition (i.e. ordering, building, delivering), time for installation, repairs and preventative maintenance, as well as restricting the scale down of assets to coincide with depreciation schedules. Furthermore, private clouds are typically built to handle peak capacity and more (i.e. N+1 or 2N configuration). The result is that during normal usage periods resources are underutilized. With public cloud providers providing seemingly infinite resources, it’s easy to see how a company could implement automated scaling triggered by certain events (e.g. performance, capacity, peak user activity). Essentially, a company can build for minimum capacity and scale up only during peak times, scaling down when peak usage subsides.

In addition, public clouds provide the means to enable self-healing infrastructure. When abnormal state conditions are detected, a replacement is spun up and the misbehaving resource is spun down.

Measured Service

Businesses with private clouds have two challenges when it comes to measuring services: how to charge business units for the use of the service and who to charge for the unused capacity. Public cloud providers keep it simple and follow the utility model: pay only for what you use.


Why is the Hybrid Model So Popular?

Then there’s the hybrid model: public-private. Looking at the large public cloud providers, it’s easy to see how they meet the essential characteristics through economies of scale, the appearance of infinite resources and the elimination of the physical layer. Why then is the public-private hybrid model so popular?

  1. Hardware vendors are not in a hurry to reduce their customer lists to businesses that are providing public cloud services. So, they have a vested interest in keeping your business tied to a private cloud.
  2. Medium and large scale businesses tend to be more comfortable with a slower, steady transformation to public cloud services. Besides the FUD factor, businesses are not going to walk away from significant capital investments. With a well-architected hybrid model, businesses can easily leverage the public cloud for “burstable infrastructure” while reducing their private cloud investments over time.
  3. Technology staff are resistant to the public cloud. At one of the conferences I attended, Glenn O’Donnell, Vice President and Research Director at Forrester Research said: “If you’ve got Administrator in your job title, you’re a dinosaur.” There are many hardware administrators and data center managers that are resistant to redefining their job responsibilities at the risk of eliminating their own jobs because they don’t view the public cloud as an opportunity for personal growth.

Many of the public cloud providers recognize these barriers and are offering private cloud options based on their public cloud architecture. Which brings me to another question, how soon will it be before these public cloud providers start leveraging your excess private cloud capacity as an extension for their public clouds?