private cloud vs. public cloud

Private Cloud vs. Public Cloud

Private Cloud vs. Public Cloud

Why hybrid cloud is looking like the model of the future

Google’s recent massive outage of services like Gmail, YouTube, and others reinvigorated an ongoing debate about public cloud vs. private cloud (also referred to as on-prem) versus hybrid cloud. This post will explore the pros and cons of each. For instances where the answer is “it depends”, we’ll look at where one might be favorable over another.


The debate usually revolves around three main factors: reliability, scalability, and cost. We’ll focus on those three. For the sake of brevity, we’re rolling security and control into reliability.

First, some quick definitions:


Public cloud: Widely available cloud services resources, like Microsoft Azure, Google Cloud Platform, Amazon Web Services (AWS), and others. 

Private cloud: Proprietary clouds, often set up by companies or other institutions for their own private use. While private cloud is sometimes called “on-prem”, or “on-premises”, that name is somewhat misleading because the equipment doesn’t necessarily have to be on the company’s premises. Often, it’s located in a multi-tenant (colocation) data center.

Hybrid cloud: A mix of public and private elements.

IaaS vs. PaaS vs. SaaS: Infrastructure/Platform/Software, as a service. Everything’s available “as-a-service” these days. We even talk about offering data center “partnership as a service“. Essentially, as you move from IaaS towards SaaS, the provider is responsible for a greater percentage of the total stack.  

public cloud vs. private cloud
Source: Nutanix


As mentioned above, we’re rolling security and control into reliability. Since the owner of a private cloud has full control over both the hardware and software environment, the degree of security, and reliability more broadly, are largely up to the owner. We should note that the owner will have more control over the hardware and the software (and even the building itself) if they have a lease in a wholesale data center rather than an SLA with a managed services retail colocation facility.

On the other hand, while public clouds are generally very reliable, there are notable failures, such as this week’s massive Google outage. Zoom, AWS, and Microsoft Azure all had big outages in 2020 as well. 

private cloud vs. public cloud

Google Cloud Platform’s SLA aims for 99.95% uptime. While that may sound great on the surface, that translates to the following average downtime:

  • Daily: 43s
  • Weekly: 5m 2s
  • Monthly: 21m 54s
  • Quarterly: 1h 5m 44s
  • Yearly: 4h 22m 58s


It’s easy to imagine many applications that would not be able to accept nearly 22 minutes of average downtime per month.

To be fair, major public cloud platforms often over-deliver beyond their SLA, but it isn’t guaranteed. On the other hand, a private cloud can be designed to deliver nearly any uptime goal. It’s understood however, that the resources required to deliver greater uptime past 99% isn’t a linear relationship. It’s exponential. Each incremental improvement in uptime requires more resources (and perhaps more usability compromises) than the last incremental improvement.

Security is similar. A private cloud can be built to be nearly impregnable, but public clouds have more vulnerabilities by their very nature. As with uptime, a private cloud can be designed to virtually any security standard, but with exponentially increasing costs in terms of resources required and usability. 

Highly regulated industries, like finance and healthcare, have restrictions about what kinds of data can be stored in public clouds versus private clouds, primarily due to security/privacy concerns. There are also data sovereignty issues to keep in mind. With consumer privacy laws, like GDPR, companies can be forced to keep certain kinds of data in certain geographies. That may limit the use of public cloud. 


Generally speaking, this is where public cloud shines. The allure of public cloud is that it is far more easily scalable than private cloud. Especially on short notice. During demand surges, such as the broad initial COVID stay-at-home order in March 2020, public cloud can quickly and seamlessly scale to meet demand.

Private cloud takes more time (and CapEx) to scale up because the capacity has to be built up.

Hybrid can be a solid compromise because an institution can build their private cloud aspects to meet the projected load, while reserving public cloud for bursts of demand. 


Public cloud costs can escalate quickly, especially during periods of surging demand. A decent analogy is time-of-use electricity pricing from your local power utility. In a time-of-use structure, energy becomes more expensive during peak usage times. There might also be tiers, where a high-consumption home is billed more when it breaks into a higher tier. A public cloud might similarly charge you more as you break into new consumptions tiers, and/or during periods of generally elevated demand.

This lack of cost control is one factor that drives many institutions towards private or hybrid clouds. 

One of the more famous cases of “cloud repatriation” – when a company rebalances load back towards private from public – is Dropbox. In 2016, they moved most of their data from AWS to their own colocation resources, which, according to their S-1, saved them just under $75 million in OpEx over two years. There are plenty more examples however. Facebook has put most of the data load from both Instagram and WhatsApp into its own data centers.

For private cloud, costs are fully within your control. Since most of the costs involved are up-front CapEx costs, you can amortize those costs over the life of the equipment and buildings.  

With a hybrid cloud model, an institution can gain more control over costs by limiting the usage of public cloud to burst demand, while building and amortizing private cloud resources for the forecasted load.

The accounting difference we’ve already shown here is that public cloud costs are essentially treated like operating expenses, whereas costs involved with building out your private cloud are treated like capital expenses. Depending on your business, one accounting treatment may be preferred over another.

How Should I Decide?

The macro trend in the business world is towards hybrid cloud. There aren’t many companies that will need only public cloud or only private cloud. It’s our hypothesis that this trend will accelerate in the foreseeable future. 

The scale of your business impacts your decision greatly. Large institutions with big data loads will be able to reach economic efficiencies of scale for their private cloud investments. Small and mid-size institutions will find it more difficult to achieve those economies of scale. 

The industry you’re in clearly impacts your decision. With this week’s news that Russia has hacked far more deeply into U.S. government systems than the U.S. leadership realized, we’ll likely see government IT loads trending towards more secure forms of private cloud. Finance and healthcare will also index higher in private cloud than most other industries.

The nature of your compute and storage workloads also deeply impacts your decision. The emerging trend towards edge computing is increasing the attractiveness of a hybrid solution. Edge computing essentially entails bringing the compute and storage closer to the point of usage. If latency must be very low, a local private cloud might be necessary. 

Your disaster recovery strategy will also impact your decision. You may back up private cloud loads to public cloud, or vice versa. On the other hand, you might back up private with private, or public with public. 

To maximize control over their private cloud, large enterprises might even consider taking an ownership stake in their multi-tenant data center (MTDC). This provides a level of control that a standard colocation lease or SLA can’t provide. 

About Prime Data Centers

Prime Data Centers develops, acquires, and operates data centers for large enterprises and hyperscalers. Through our own properties, customer assets (sale-leaseback), or greenfield developments, Prime delivers build-to-suit, powered-shell and turnkey solutions. A private firm owned by a group controlling $6 billion of assets with a 15-year tenure in technology and real estate, Prime provides customers with ownership options and dynamic leasing models defining a true corporate partnership. 


Get our once/month news delivered to you:

we can advise your data center development

Chat with us today. We're happy to provide strategic advice.

Contact Us


4701 S. SANTA FE

33 MW of critical capacity in Vernon, CA.

Los Angeles Data Center

Well-connected, hyperscale capacity minutes from downtown.




Are you the next Prime-Time Player?


Join a growing company in an in-demand industry.