Shifting to Software-Defined
Showing results for 
Search instead for 
Did you mean: 

Why the Public Cloud is NOT For You!





Why The Public Cloud is NOT for You!


Hybrid Cloud.jpgHybrid CloudWhat do you mean that the public cloud is not for me? Are you not aware that public cloud spending is growing around 20% annually? All my competitors are moving to the public cloud. “No” is not an answer.

Yes, the public cloud is here, growing rapidly, and can support many of your workloads. Keep reading and I believe you will convince yourself that the right answer is a hybrid cloud, defined by Gartner as, “Hybrid cloud computing refers to policy-based and coordinated service provisioning, use and management across a mixture of internal and external cloud services.” The internal (private) cloud can be run on premise, off-premise, or managed on or off-premise by a third party. You will not be alone in deciding on a hybrid cloud. Ninety percent (90%) of enterprises say they are going to pursue a hybrid cloud solution this year.[1]

I stumbled upon this analogy while building a business case to use the Open Compute Project (generic) servers for a customer’s workloads. This analogy may help you get around the gut response that public cloud is always the answer. Southwest Airlines (SWA) is the largest domestic US airline. Their business model, in their words is, “Southwest Airlines was built on providing consistently low fares with superior Customer Service while maintaining a low-cost discipline. We’re proud to have never declared bankruptcy or furloughed Employees. We do things differently than our competitors …...” Southwest continues to be profitable year-after-year in what is a very cyclical industry. What is their secret? One is their low-cost discipline that allows them to provide equivalent services at lower cost than their competitors. SWA can make money at prices where their competitors lose money. SWA has the lowest costs in the industry on their routes, achieved largely through their use of a common fleet of aircraft. SWA runs the world’s largest fleet of Boeing aircraft, all 737s. This allows SWA to train pilots, flight attendants, mechanics, ground crews, and gate agents on one airplane; and all employees can support any airplane in their fleet. This sounds similar to large public cloud providers using common X86-based servers across their datacenters. However, with commonality come challenges.

SWA, with their one aircraft type, tailors their business to the capabilities of their airplanes, primarily range and capacity. SWA is an international airline, with service in the US, Mexico, and the Caribbean. The 737s lack the range for service to South America, and are not certified for trans-Atlantic and trans-Pacific flights. SWA could fly to Canada, but an estimated 75% of Canadians live within 100 miles of the US border, so SWA can serve them from border US cities. SWA is also restricted in how small they can go, as a route needs to provide a reasonable load factor in a Boeing 737 (i.e., 137-seat and 175-seat capacities). It is common for other large carriers to partner with regional airlines that fly 50-seat and smaller planes to serve small and tiny markets. Summarizing, SWA has the lowest costs in the industry by using one aircraft type and tailors their service (workload) to this one aircraft type.

Using this analogy in public cloud, a single generic server type (i.e., X86 server) is the right answer if all ones workloads are similar and run efficiently on these servers. It is a low-cost acquisition model for public cloud providers that provides an economy of scale that private clouds probably cannot match. However, what happens for those workloads that do not run efficiently on the generic cloud servers? You, much like SWA wanting to enter a market not suited to their fleet, have unattractive options. You can pay for unused capacity (e.g., 137-seat aircraft for 30 passengers), overcome inefficiencies with more servers and infrastructure (e.g., second 737 to customers 138-150 in a market), or have your workloads run slower on sub-optimal servers (e.g., inadequate facilities lengthening turnaround time on the ground). Switching to another public cloud provider will likely not change the economics as most cloud providers have similar generic servers and infrastructure. Your most effective solution is a hybrid cloud with your private cloud running your non-standard workloads, whether hosted on-site, off-site, or off-site and managed by a third party, and the public cloud for generic workloads where you can benefit from their economies of scale and desire to run razor-thin or negative margins. Your private cloud consists of purpose-built servers and composable infrastructure

Let us take a moment to understand the approach used by “cloud giants” in specifying and purchasing their generic servers and some of the costs beyond server acquisition costs. You will understand the pros and cons of their approach for your workloads. From this understanding, you can determine what factors are important to you in choosing where to run your workloads in your hybrid cloud.

Cloud giants run over 1 million servers in multiple data centers, both for their workloads and their customers’ workloads. Their infrastructure is so important that they design their own infrastructure, buy their own dark fiber, and are a full-service provider. They optimize their infrastructure to their datacenters and applications. They rewrite applications to run on their infrastructure. Their operations and maintenance team is trained for their infrastructure and only their infrastructure. They do have low purchase prices, but so does anyone who buys large quantities of anything. Using the SWA analogy again, SWA’s fleet supports a variety of markets and flights, and SWA matches their services to their fleet. SWA does not fly into very small markets, or on trans-oceanic flights. SWA would have large initial costs to introduce a new aircraft into their fleet to support a new service, either a very small market or a trans-oceanic flight. To date, the potential additional revenue has not justified the large initial cost to add additional aircraft types to their fleet. It was after much deliberation that SWA decided to purchase some 737-800s with 175-seat capacity to serve new airports (routes acquired in their acquisition of AirTran) that had limited landing slots. SWA continues to expand into markets that they can satisfy with their current aircraft. The cloud giants do not have the competitive concerns of SWA (someone’s aircraft being better for a specific type of market), as the cloud providers’ competitors run similar large fleets of generic servers. You, as a public cloud provider customer, will either pay for the inefficiencies in their public cloud, or maintain a private cloud as part of a hybrid cloud.

IT departments are in a different place than SWA. SWA started as a small Texas airline, purchased 737s, and expanded into markets where they could maintain low costs with their 737s. IT departments are established and have a diverse set of workloads running on a variety of servers. While some of these workloads are suitable for both the cloud and run efficiently on generic servers, many are not. Forcing yourself to run these workloads on generic servers is analogous to SWA forcing their way into new markets and services with their existing fleet. SWA could make that change, but not while maintaining their low-cost structure.

The public cloud surrounds us, but remains difficult to define and even harder to quantify. Cloud providers hold regional and nationwide events and summits where customers and potential customers are in awe of the possibilities, but not able to quantify the tangible and intangible costs.

How does one determine the best mix of private, hybrid, and public cloud? The figure below shows an approach based on understanding your workloads, defining policy and architecture standards (e.g., compliance, disaster recovery, geographic), planning, and migrating workloads based on justifications. Following the figure, several methods for justifying the migration from legacy to private, hybrid, and public cloud are detailed.


Journey to Hybrid IT.jpg

Journey to Hybrid IT[2]

Costs. The stock answer is to consider more than acquisition costs and look at Total Cost of Ownership (TCO). “Gartner defines total cost of ownership (TCO) a comprehensive assessment of information technology (IT) or other costs across enterprise boundaries over time. For IT, TCO includes hardware and software acquisition, management and support, communications, end-user expenses and the opportunity cost of downtime, training and other productivity losses.”[3] Calculating TCO is challenging (impossible?) as one needs a wide range of estimates for costs other than initial purchase prices. Hardware replacement costs can only be forecast at the time of calculating TCO. One also has a concern of vendor (cloud) lock-in where an initial choice of public cloud provider is difficult-to-impossible to later change. Hewlett Packard Enterprise’s Eucalyptus software lessens this cloud lock-in with you now able to pull workloads from the public cloud to a private cloud.

Workload. An alternative approach is to characterize each workload (performance per $ of infrastructure investment) on both commodity and purpose-built servers. This is challenging due to the diversity of workload until one manages to rewrite all applications to run on top of a single distributed compute model. Another trap in this approach is one can have many server solutions, each running a low utilization level as diverse workloads cannot run across uncommon servers. Characterizing each workload becomes more complex when one analyzes beyond X86 processors and factors in other models of compute including GPUs, Systems on Chip (SOCs), etc.

Cloud Brokerage Enablement Calculator. Another approach is cost-based and uses a cloud brokerage enablement calculator like CloudGenera’s CloudAssist™. This calculator lets you configure your environment or applications, and see whether on premise, private cloud, or public cloud, and which public cloud, best meet your needs. Best fit to your needs can include cost, compliance (e.g., HIPPA), security, or other factors. You can evaluate how changes in your configurations affect costs.

Similar to the holy grail of moving to the public cloud, some companies are looking to purchase a white-label infrastructure (i.e., generic X86 servers, networks, storage) to get the lowest acquisition costs. One rationale is that they will have a single infrastructure to support and operate. This is not true, as server lifespans are typically 3-5 years, and new generations of servers come out every 12-18 months. Add in that software upgrades occur and software goes EOL (e.g., Windows Server 2003 on July 14, 2015) requiring operations teams to upgrade. The reality is that purchasing generic infrastructure still results in supporting at least two generations of hardware while deploying a third generation with the higher number of combinations of old with new. Non-generic servers have the operations challenges of projecting vendor stability including the vendor’s ability to manage a benchmarks roadmap, supply chain stability and security, unique support requirements, etc.

This is not a simple problem to solve. Given that, do you really believe that one size (public cloud) fits all?




[1] Logicworks.

[2] Hewlett Packard Enterprise

[3], Total Cost of Ownership (TCO)

About the Author


Technology entrepreneur with 30+ years experience successfully launching new products and services for small-to-large companies. Now applying his knowledge and experience at Hewlett Packard Enterprise as a Chief Technologist. Also active in the Seattle start-up community. Educated in Physics at MIT and has an MBA from the Wharton School, University of Pennsylvania.


I read an interesting post on another site opining that the white box servers (generic servers) are turning the tide back toward private cloud with private cloud getting cheaper for the smaller guys and economy of scale is going away for the CSPs.


See posts for
HPE at 2018 Technology Events
Learn about the technology events where Hewlett Packard Enterprise will have a presence in 2018.
Read more
See posts for dates/locations
Reimagine 2018
Join us at one of the Reimagine 2018 stops and see how we Simplify Hybrid IT, innovate at the Intelligent Edge and bring it all together with HPE Poin...
Read more
View all