Public cloud computing encompasses a range of options, from IaaS VMs to container services and serverless. All of them have their place, but serverless computing has significantly grown according to some reports.
Although serverless is reaching maturity, it’s not sweeping the cloud market. Less than 20% of public cloud users said they use serverless computing, according to Andover Intel, a consulting and analysis firm. Cloud overall is increasingly seen as a partner or hybrid strategy rather than the inevitable destination of all computing. Similarly, enterprises might consider serverless as a partner cloud strategy, something used with other cloud services, particularly container services.
How are enterprises using serverless, and how can we reconcile these data points? If they can’t be reconciled, what’s the true state of serverless?
Serverless rewards and risks
The principle behind serverless computing is simple. Traditional VM or container services require users to commit and manage persistent resources, then wait for work to be presented. In contrast, serverless computing loads application components dynamically when needed, so teams don’t need to manage resources.
Enterprises realize the benefits of serverless in its dynamism and scalability, not in the elimination of resource management. If work is sparse, serverless doesn’t waste money paying for unused resources. If workloads are highly variable, the same is true because it’s not necessary to commit persistent cloud resources to handle occasional peak loads.
Serverless computing is arguably aimed at event-driven applications, including IoT, which represents a fast-growing segment of enterprise applications. Over 90% of the enterprises who reported serverless usage said their applications were event-driven, most classified as IoT. This demonstrates that serverless benefits are more specialized.
As such, enterprises must balance the benefits of serverless with certain risk factors, such as the following.
Over 90% of the enterprises who reported serverless usage said their applications were event-driven, most classified as IoT.
Cost overruns
Generally, the unit cost of computing in the cloud is higher for serverless than for containers or VMs. As the amount of presented work events grows, serverless costs mount proportionally and eventually exceed traditional cloud service costs. It’s often difficult to predict the usage of applications, and usage underestimation can easily make serverless too expensive.
Latency
Because serverless application components load and run dynamically, delays associated with their execution might affect application performance. If several serverless applications are chained together, the delays are additive.
Requires stateless programming
Because a serverless component doesn’t stay resident between uses, it’s not possible to store data in one. In cases where development teams process events partly based on state — their context relative to other events — they manage state information another way. This requires a different application programming model than most development teams are used to.
Developments in serverless
In 2023, important changes helped enterprises address the risk and reward balance of serverless. These, in addition to the increased interest in event-driven applications, are responsible for the growth in serverless usage. These changes have not yet affected all cloud users, which is largely why serverless still accounts for only a minority of cloud usage. Some of the main changes with serverless include the following.
Messages recognized as an event
Most of IT focuses on transaction processing, which involves accessing and updating records of business activity, such as receiving and making payments, ordering and processing orders, and business planning and analytics.
Enterprises are now realizing they can treat messages, such as emails, texts and calls, as events and handle them via serverless computing. Texts-as-events is currently spreading serverless to more enterprise users than IoT.
Increased developer experience
Developer experience and specialized tool inventories for functional or microservice development continue to grow. Most developers now understand serverless and stateless components. Public cloud providers have extensive documentation on serverless development, including tutorials.
Additionally, the availability of cloud provider and third-party tools to facilitate development and application lifecycle management has burgeoned. Just under 80% of enterprises told Andover Intel their development teams could handle serverless projects, up from only 18% in 2018.
Hybrid applications
An emerging trend is to create a hybrid application that includes serverless and another compute model. The initial hype about serverless focused on the elimination of server management as a goal. This simplistic proposition led enterprises to discount applications not suitable for totally serverless implementations and fail with applications that careful review would have disqualified.
Now enterprises understand that serverless computing is more likely to be a piece of an application than the entire one. The majority of serverless computing in new applications is part of a hybrid compute model, not an exclusively serverless deployment.
Better planning
The majority of serverless computing in new applications is part of a hybrid compute model, not an exclusively serverless deployment.
Enterprises are reviewing their cloud planning processes and past decisions. The notion of repatriation, or returning applications from the cloud to a data center, is a small part of this development. Enterprises have recognized that the cloud has features and benefits that dictate its optimum use and not everything needs to move to a cloud environment. This realization has also guided an awareness of what serverless is well-suited for and how to realize and preserve its benefits.
The cloud and serverless computing have been used in ways that don’t fulfill their key value propositions. But a clear relationship exists between serverless and cloud containers. Both are dynamic and elastic, and both offer a different trade of performance and latency against costs. Some container applications will partially or completely transform into serverless, while some serverless applications move the other way.
The state of serverless in the current environment is defined by a new realism, and that’s a good thing.
Tom Nolle is founder and principal analyst at Andover Intel, a consulting and analysis firm that looks at evolving technologies and applications first from the perspective of the buyer and the buyer’s needs. By background, Nolle is a programmer, software architect, and manager of software and network products. He has provided consulting services and technology analysis for decades.