Microservices / Observability / Serverless

Solving Serverless and Tracing Is Key to Success in Observability

14 Jan 2020 12:39pm, by

Using logs, metrics, and traces together is part of the definition of “observability,” according to 83% of the nearly 1,000 respondents to Logz.io’s annual DevOps Pulse study. Since the survey question had the option of “select all,” another finding is that two-thirds said observability is a measure of how well a system’s state can be inferred from output data. That latter definition is the one preferred by observability pioneer Charity Majors in her “Observability — A 3-Year Retrospective” post. End-users’ impressions of what these terms mean are important to Logz.io because the company itself is a cloud observability platform built on top of the ELK stack.

The vague term “monitoring” is considered critical to achieving observability, but so is alerting, tracing, instrumentation and anomaly detection. Although these practices are prerequisites, companies have been slow to adopt specific tooling. Notably, only 19% are using a distributed tracing tooling like Jaeger. Whether or not this is the “correct tool” is an open question, but finding the right tool for the job is an observability challenge for more than a third of the study.

A bigger challenge is complex architecture, cited by 58%, which can be associated with microservices and serverless trends. In fact, in a different question serverless was named ahead of service meshes, containers and orchestrators as an obstacle to achieving observability.

Readers of The New Stack may remember that in 2018 our “Guide to Serverless Technologies” we reported 60% of people using serverless architecture said debugging and tracing was the biggest monitoring pain point. The survey also found that among those hadn’t started using serverless, debugging (50%) and monitoring (35%) was expected to be the top challenge when deploying and managing serverless architecture. Hidden within the Logz.io report, we found that the adoption of serverless into application logic actually dropped from 43% in 2018 to 42% in 2019. Perhaps this is a temporary respite in serverless growth, but it also may be a warning sign.

We plan to dig into other parts of the 2019 DevOps Pulse in the future. In the meantime, here is our take on one more chart:

What architecture are you using to build your application?

Only 19% of the study was only using monolithic architecture, but that figure dropped to 7% this year. The percentage of those only using microservices architectures rose every year and now represents 37% of the study. Since complex architecture is the top challenge to achieve observability, this might explain why developers started to shy away from monolith/microservices hybrids in 2019.

Feature image via Pixabay.