Serverless computing for CFOs

May 24, 2019

A blog post by Eric Lam, senior manager, Deloitte Consulting and Karthik Pallapothu, senior consultant, Deloitte Consulting

The last decade of cloud computing advances have helped IT transcend old infrastructure constraints and develop and deploy software at scale while also driving down operational costs and improving time to market. Now, cloud computing is moving toward even greater flexibility and smaller units of scale in the form of serverless computing.

Serverless computing has begun to create major buzz among executives charged with making enterprise IT decisions. With its inherent ability to create cost savings and other potential benefits, IT teams and CFOs are taking note and may be tempted to rush into a serverless strategy. As with any major shift in IT, there are many factors CFOs must consider to ensure a serverless approach is right for the enterprise, and employ it in a way that delivers the desired ROI.

Serverless computing: how it works, and why CFOs should care

“Serverless” doesn’t mean “no servers.” In the traditional cloud computing model, enterprises pay a fixed, recurring amount for using virtual servers to run properties such as websites or applications, regardless of whether the cloud services were in use or not. In contrast, serverless computing allows the enterprise to pay only for actual usage and avoid paying for idle time.

As the application load increases, more instances of the application are spun up proportionately, and scaling of the underlying infrastructure is managed by the cloud provider. This represents a critical shift in burden: the enterprise does not need to drain resources on planning for scaling or on capacity planning and management or the countless tasks associated with maintaining virtual machines, server farms and operating systems.

Lookouts for CFOs: hidden impact on budgets, billing and overall ROI

Serverless implementations naturally change the way costs are calculated and how organizations are billed. While they offer huge cost-reduction potential and increase productivity, they also pose capital planning and operational challenges for organizations that are not adequately prepared to go the serverless route.

• Unexpected cost spikes during periods of high demands
  Serverless platforms come with the inherent capability to handle infrequent or unpredictable application demand and scale resources accordingly. In a traditional on-premise architecture, a retailer might procure additional servers in anticipation of heavy demand in a particular week. The retailer could forecast costs for the additional physical servers with fairly high accuracy. However in the serverless model, the system essentially running on auto pilot. Pricing is determined by the length of time an application runs and the number of times it is called. Without proper system monitoring and safeguards in place, costs can quickly balloon.
• Hidden costs from micro delays and seemingly minor code changes
  Serverless computing instances are only active for the duration that an application or component runs. If a serverless instance is called after a period of inactivity, there can be a slight delay known as a “cold start” while the inactive instance warms up before it can run. This increases the total billable runtime and must be considered when determining ROI. Developers should also be aware of code changes that could rack up billing costs. For instance, if a small change in application code increases the runtime of an application from 100 milliseconds to 200 milliseconds, the cost doubles. If the application runs thousands of times per hour, those milliseconds can add up quickly.
• Potential accounting complexities
  The pay-per-execution model of serverless computing can be a challenge in terms of accurate accounting. In organizations with a shared service model, serverless applications should be designed to tag traffic and map usage back to the respective lines of business either through chargebacks or showbacks.
• Refactoring of legacy applications
  Serverless architecture can vastly improve efficiency by breaking monolithic systems into smaller, modular components designed for specific functionalities. In order to fully and accurately estimate conversion costs and ROI, it’s essential to consider the cost of refactoring of any mission-critical legacy applications. In some cases, the added cost of refactoring and reskilling developers to support the new architecture may run counter to business objectives or outweigh the advantages offered by serverless computing.

Making the right call for your organization

Even with serverless computing still in the very early stages of maturity, it is already helping organizations bring products to market faster, shorten innovation cycles, increase agility and productivity, and streamline IT operations generating potential cost savings with its consumption-based model.

However, just as with any technology decision, it’s critical to consider the unique challenges and use cases of your business, map out requirements and conduct disciplined implementation planning. While serverless computing will not be the right solution for every business every time, a hybrid solution may be able to help your business achieve key business objectives and demonstrate measurable results. With careful due diligence it is possible to move forward confidently with a serverless solution that meets or even exceeds the desired ROI for your organization.

Interested in learning more? Check out our comprehensive report, Serverless computing architectural considerations & principles released in 2018.