Developer Empowerment and Cost Optimization Drive Serverless Computing Adoption

As cloud has become part of everyday IT discussions in enterprises, many IT managers are asking, "What's next?"

According to Gartner analyst Arun Chandrasekaran at the Gartner Symposium/ITxpo in Orlando, Florida, in October 2018, software innovation and the strengthening of developer empowerment in enterprises have made serverless computing one of the "hottest" areas in computing.

Chandrasekaran noted that developers spend about 30% to 40% of their time on infrastructure-related work or waiting, and serverless computing allows them to shift their attention to other matters. Serverless computing acts as a bridge between Infrastructure as a Service (IaaS) and Platform as a Service (PaaS), enabling users to be unaware of operations and management, and supporting pay-as-you-go and on-demand scaling.

For most enterprises, serverless computing is still in its early stages, and enterprises need to weigh the advantages of this computing platform against other abstraction layers such as containers and virtual machines. Early adopters have reaped many rewards, but misunderstandings about the technology's nature and value remain widespread in the market.

Chandrasekaran said the technology's name is still misleading and is often confused with Platform as a Service. In reality, operational tasks such as provisioning and management are completely abstracted from users, but servers are still used in the background.

Furthermore, serverless computing is not limited to compute: it resembles quasi-PaaS, provisioning beyond just the underlying compute capabilities. Its difference from PaaS is that in a PaaS environment, users are still responsible for scaling, whereas in serverless computing, scaling is abstracted away.

Current state of technology

Convincing developers to invest in serverless computing can be quite challenging.

Tom McLaughlin, founder of ServerlessOps, told CIO Dive in an interview that people working on serverless computing are passionate about the field, but the challenge for many vendors is that the rest of the market "doesn't really care" and is not yet "particularly impressed or fascinated."

According to forecasts, the market size will reach $7.7 billion in the next three years, and major cloud vendors are vying for the leading position. By some estimates, Amazon Web Services (AWS) holds a dominant 58% market share, but competitors are catching up quickly.

Google Cloud Platform's serverless Cloud Functions platform has grown particularly rapidly, with customer preference rates exceeding Microsoft Azure Functions at 23% and 10%, respectively, according to CB Insights data.

IBM Cloud Functions is another popular platform in both public and private clouds. Top private cloud platforms also include Azure Functions Runtime, Nuclio, OpenFaaS, and Knative, but Chandrasekaran noted that the full benefits of serverless computing cannot be fully realized in on-premises environments.

Michael Behrendt, IBM Distinguished Engineer and Chief Architect for Serverless and Event-Driven Applications, told CIO Dive in an interview that enterprises finding valuable applications for serverless computing are reaping the benefits of resource and cost optimization. The technology helps customers use data in unprecedented ways and build new pathways to support customers.

But successful implementation also requires mitigating friction that arises during the implementation of new technologies.

McLaughlin believes that disruptions will occur in operations, but such disruptions can be a good thing. Serverless breaks the "lift and shift" model and helps achieve DevOps.

Many companies that have not fully embraced cloud or DevOps are "leapfrogging" to serverless computing. He said that migrating to the public cloud can take years, and companies may fall behind again after three to five years. Therefore, some customers go directly to containers or serverless.

Evolution of compute abstraction layers

Customers are not faced with a strict trade-off between on-premises, cloud, and serverless computing. There is a spectrum of compute abstraction layers, each suited to specific workloads.

Chandrasekaran pointed out that physical servers are an appropriate choice when the business requires low latency, deterministic performance, or isolated workloads (the latter often due to security, compliance, or licensing requirements).

Although physical servers remain popular for legacy applications, edge computing, high-performance computing, and big data applications, their demand is "diminishing," and many such applications will skip a generation and move directly to containerized or serverless environments.

Chandrasekaran said that hypervisors (i.e., virtual machines) are the backbone of many enterprise workloads, but they often suffer from significant sprawl and high total cost of ownership, especially maintenance costs.

Containers and serverless offer more operational control, but quality is not as "built-in" as with hypervisors.

Chandrasekaran noted that containers are popular in testing and development because code can run consistently without needing to be written multiple times. Containers are suitable for legacy workloads, cloud-native applications, and per-node licensing. They can start quickly to handle agile workloads, but they are still immature, especially in enterprise features.

The advantages of serverless computing lie in pay-as-you-go cost efficiency, scalability, operational simplicity, efficient resource utilization, and increased developer productivity. It also integrates well with microservices architectures.

But it is not perfect: Chandrasekaran pointed out challenges such as latency, vendor lock-in risks, insufficient ecosystem support and platform capabilities, and weaker debugging and monitoring.

Behrendt said that serverless computing focuses on the rapid execution of stateless, self-contained units of logic, so users must also accept the operating system provided by the platform. If a user needs to use the Windows operating system, but the serverless platform provides Linux, then the application is not suitable, and the user may have to downgrade to a container or virtual machine platform.

It comes down to a trade-off between convenience and control: a platform where users don't need to worry about many things versus the control offered by lower levels.

Future direction

Serverless computing usage has grown significantly over the past two years, but the technology is still in its early stages of enterprise adoption.

Behrendt believes that in the coming year, serverless computing will expand its scope to be applicable to more workloads than it is today, and shortcomings in areas such as runtime and memory will be mitigated or disappear entirely.

McLaughlin said that as the field develops, more companies will focus on reducing developer-oriented products, which will help change the perception and discussion surrounding the technology.

CIOs and technology managers should seriously discuss what capabilities and advantages serverless computing can bring to their organizations, and whether the investment risks are justified.

Behrendt advises: "Start small, but start fast." First gain experience with a small function, then gradually expand the scope of applications using serverless. There are already many precedents of customers running workloads on serverless computing, so there's no need to start from scratch.

Enterprises also need to provide training in cloud IaaS and PaaS environments and ensure developers understand cloud-native design principles.

Chandrasekaran pointed out that cloud-native applications are stateless, with services loosely coupled, so application units are independent. This means that in containers and serverless computing, developers focus on instances rather than infrastructure.

Enterprises need to identify event-driven use cases with isolation and independent applications. If already in a virtual machine environment, identifying workloads that can be migrated can bring immediate cost savings.