It’s called serverless computing, introduced with services such as AWS Lambda and now Microsoft’s Azure Functions. These systems manage the starting and stopping of machines for you. They’re not really serverless, but the servers aren’t yours to buy or maintain. You’re billed as an abstract measure of the resources you used, not for the runtime of those servers.
The single most popular question I now get: What are the upsides and downsides of serverless computing? Here’s my take.
Upsides of serverless computing
First of all: Cloud providers, what took you so long? This always seemed like a better way to manage server resources. With serverless computing, I don’t guess any longer as to what servers I will need to run my work—which is dynamic and changing. Serverless computing deals with that on my behalf.
Second, serverless computing should cost less in the long run. The cloud provider will manage the resources you need, and it won’t allocate resources that you don’t need—hopefully. Thus, the bill should be lower if both you and the cloud provider are doing your jobs correctly.
Downsides of serverless computing
Serverless computing is a great ida. But it’s not perfect.
Performance can be an issue. The model itself means you’ll get greater latency in how the compute resources respond to the requirements of the applications. If performance is a requirement, it’s better instead to use allocated virtual servers.
Monitoring and debugging of serverless computing is also tricky. The fact that you’re not using a single server resource makes both activities very difficult. (The good news is that tools will eventually arrive to better handle monitoring and debugging in serverless environments.)
Verdict: The upsides outweigh the downsides
Should you use serverless computing? Yes. Count on Amazon Web Services, Google, and Microsoft to focus more on their serverless offerings. AWS recently provided pretty good updates for Lambda, for example. More are on the way.
At the end of the day, serverless computing is a better way to consume cloud resources. It’s not a different paradigm, but it’s been a long time coming.