Serverless Computing | Top Considerations

Serverless Computing | Part 20 of AFCOM’s Consider This Series

Serverless computing is a way to provide an abstraction layer between server resources and the functions they provide. It is also called Function-as-a-Service (FaaS). In essence, the functions of the server are provided via the cloud. Serverless computing, therefore, is a bit of a misnomer. Servers are still involved, but they’ve been moved from the data center to the cloud provider.

Here are the top five considerations for data center managers considering serverless computing:

Consider This:

1. Offloading Server Admin

These days, data centers are often understaffed, which has led some to consider the duties that are most essential to maintain in the data center. For some, it might be application development or hosting of storage or databases. For those where server administration is a distraction from core duties, offloading server duties can free up a lot of time.

Serverless indicates a way to build and run applications and services without the need to manage the underlying server infrastructure. The servers are dealt with by a cloud provider. That provider allocates compute resources in response to user demands. Data center personnel no longer need to spend time on functions such as provisioning, scaling, and management of server infrastructure.

2. Application Development

For those operations where application development and testing are major tasks, serverless will have plenty of appeal. It frees up developers so they can focus on coding. No more fiddling about with equipment maintenance, configuration, capacity planning, fault tolerance, or the management and scaling of containers, physical servers, and virtual machines. This ultimately means that applications can be built faster. Serverless computing services eliminate the need to manage the underlying infrastructure.

3. Ending Procurement Hassles

With supply chains currently delayed, some data center managers may feel it is easier to enlist the help of cloud providers than wait months for servers to be delivered. Consider, too, that it can take weeks or even months to evaluate, select, and obtain purchase order approval for equipment. If server resources are needed now, serverless computing opens an avenue to rapid provisioning. This raises productivity as time is spent primarily on coding, development, and other duties rather than internal plumbing.

4. Seasonal Demands

Many data centers may be more than happy to manage their own servers. They have been doing so for decades, after all. But some may look to serverless resources to satisfy seasonal demands. Data centers in retail, for example, can add servers and compute power over the holiday season. Some data centers, too, continually wrestle with the need to beef up resources based on an end of the month or end of the quarter cycle. There are also sudden spikes in demand due to the launching of a marketing campaign, the development of a new product, or sudden consumer demand.

The common solution to this short duration traffic spike problem is to overprovision resources. The data center buys enough servers, compute, and storage gear to deal with peak load. Yet that equipment is only used for a couple of days every cycle. This eats into budgets and lowers overall efficiency. Serverless offers a much cheaper alternative.

If the compute resources are only needed for a short time, why build a supporting infrastructure internally? As serverless is a pay-per-use model, it should work out cheaper to rent serverless than maintain a fixed quantity of servers. Similarly, if an application proves to be more demanding than anticipated, serverless environments can rapidly provide the extra resources. This kind of dynamic scaling often drives serverless adoption.

5. Don't Put Stateful Applications on Serverless

Be aware, however, that serverless computing doesn’t work well with stateful applications. It is the stateless use cases that are the best fit for a serverless architecture. Why? It takes too much effort to enable stateful applications to run in a serverless way and is not always feasible. Those attempting to do so run into challenges such as having to store the session state in memory between requests.

There are other applications that are a good fit. Those with non-uniform access characteristics and highly variable traffic patterns, for example, work well in serverless environments. However, data centers working in heavily regulated industries are advised to avoid serverless as compliance issues may crop up due to its use of a multi-tenant model. All it takes is one misconfiguration and data exposure or other security or compliance problems could materialize.

Please keep an eye out for the next release of “Consider This."