Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Five questions to guide a cloud vs. on premises debate


Have you projected your public cloud expenses?

Source:  Apinan/Adobe Stock
Designer: Megan Cassello

Some organizations mistakenly believe that the public cloud is always cheaper than on-premises alternatives.To calculate the financial impact of moving a workload off premises, networking, workload variability and many other components come into play.

The latency of the network, for example, affects whether the cloud vs. on premises will be more financially sustainable. Low-latency workloads that organizations can easily move back and forth, and don't need to access frequently, should go into the public cloud rather than locally storing massive amounts of data.

"Public cloud is fairly cheap storage," said Fidacaro. "If you don't need access to that data at the moment for analytics, you can just pull that data from the edge or the enterprise up into a public cloud."

Consider how the workload's utilization impacts price. Sustained utilization might be cheaper to run on premises, for example, while a bursty or lightly used workload would be better to run in the public cloud, said Lowe.

Understand the nature of the application. For example, find the percentage of memory and CPU it typically consumes, and determine the size of the cloud instance it would need based on those figures to calculate potential cloud expenses.

"If you move an application to the cloud, all of a sudden, you're creating a tremendous amount of data at that location," said Eastwood. Many cloud models include a premium for the data storage, and it's expensive to move the data back on premises because of cloud expenses such as network taxes and fees.

Assess the financial risk of a potential cloud outage. If an airline's load-balancing application goes down, for example, the entire business is affected.

"There are a lot of places where you can focus your attention to get better returns on that attention rather than some of those high-value applications, where the biggest risk is the loss of access to those applications and the potential loss of data," said Eastwood. "These are things that are very core to the business that you can't afford to disrupt."

As applications become increasingly more distributed, it's essential to understand each of the different external and internal elements of the workload, said Lowe.

"You have to understand the dependencies, interconnections and components of the applications in order to understand what can or should be moved to another location," he said.

View All Photo Stories

Join the conversation

1 comment

Send me notifications when other members comment.

Please create a username to comment.

There are a number of additional considerations that relate to behavior in organisations: 
1) Integration elements that often include different technologies associated with the specific applications being integrated. "What do you pick up when you lift and shift an application" How is routing impacted by doing this? 
2) Load balance strategies and gaining a full understanding of this in the cloud.
3) Lifting and shifting applications that span across different technologies like AIX on enterprise machines (For DB hosting - often clustered), Application servers hosted on Linux or Microsoft, Load balance strategies between application servers, and then web servers. On-Premise may be a serious consideration her due to architecture or a longer term strategy is needed to create APIs that deal with integration. This should ideally be a deliberate part of the IT strategy.

Other factors relate to affordability of support renewals or rather the lack of support renewals.  Any migration of these systems often demand the requirement for renewal and then also introduces a lot of complexity where the existing code is redundant or may require multiple of iterations of updated that further introduce risk to operations. Regardless of the fact that the latest version of a solution may be cloud enabled.

Outdated operating systems due to incompatibilities of legacy code. This relates to in-house written code where development teams have moved on and legacy OTS systems. Alternative technologies may be required to "wrapper" these systems with the end state being that these systems be replaced. 

All of these may lead to additional expenses and operational risk that should be evaluated during the decision making process...