Observability to understand serverless applications


Applications and systems are becoming more complex. Also the expectations that users have against them. Twenty years ago we could find a website under maintenance for hours and, as users, we accepted this situation and simply waited. This is now unthinkable. User experiences are marked by a demand for immediacy.

In response to this context, we find the observability that today is already a must in the systems and architectures of companies.

There are three types of data that can improve the observability of our serverless systems: logs, metrics and plots.

Datadog is a monitoring and analytics platform that helps companies improve the observability of their serverless infrastructure and applicationsincluding both Function as a Service (FaaS) and hybrid applications based on functions.

It’s important to make sure we’re getting the right data from our systems without sacrificing function performance.

To expand on all this information and see a demo with an application, we recommend you watch this interesting session by Ara Pulido, Developer Relations at Datadog:

Is your application serverless or your system observable?

The answer to that question should be: yes. Having a strong observability strategy has become critical for companies with production systems, particularly for microservice architectures.

In this post we talk about observability for function-based architectures, aspects to consider and what data and metrics are necessary to understand them.

Observability vs monitoring

We differentiate these two concepts. Monitoring is an action and observability is a quality of your system. That is, your application is observable or it is not.

Observability is the ability to answer questions about the internal state of the system based on external data that, if added, the system will provide us. We refer to three types of data: the metrics, logs and plots that will allow us to test and improve observability to understand what happens in our system.

Parameters to improve observability in serverless environments

Serverless environments are very flexible and one of their great advantages is that their functions scale to zero. What does this mean? That as long as the cloud is not used, nothing is paid and this translates into great economic savings.

Serverless architectures have their peculiarities:

  • They do not allow you to access the operating system.
  • The environment that executes your function is, in many cases, a kind of black box.
  • The execution time of your function and the memory that you have reserved for your function, has a cost.

With Datadog you can integrate logs, metrics and traces into your system to improve observability

The logs they are the most basic when we work with public cloud functions because they are very simple and transparent for the developer. A classic design pattern is the Log forwarder: it is made up of small functions that will be checking if there are new logs in the cloud system and will send them, in this case, to Datadog.

In a serverless system we are interested in the metric that the cloud or Datadog offers and they are the duration of a function, the number of errors that we have, the number of invocations (traffic) and the throttles (requests to your functions that cannot be fulfilled because it has reserved a maximum number of containers that go to execute that function).

To implement the trace In a serverless system, we must apply one of these strategies or a combination of both:

  • Use the native systems of that cloud (for example, for AWS the X-Ray system; or for Google Cloud Functions, Google Trace).
  • Use an APM library that uses the logs to create that trace and then reconnect once it’s sent to Datadog.

PUE is Datadog Managed Services Provider

PUE, as Datadog’s Managed Services Provider, implements the platform and directly manages customer environments in addition to marketing licenses within the Big Data Full Stack. In this way, PUE expands its service with Datadog in the area of ​​Platform Administration and Support, thus reducing diagnostic and troubleshooting times.

Contact us to receive more information without obligation.

Links of interest

Efficient application management thanks to the consolidation of logs and metrics with Datadog

Datadog Fullstack Extreme Monitoring

PUE Sessions, the event on Big Data and Cloud trends closes its annual edition with successful participation


training@pue.es for official training in benchmark technologies.

exams@pue.es for official certification in reference technologies.

sales@pue.es for professional services in Big Data and Cloud.


Related Posts

Leave a Reply

Your email address will not be published.