Book Image

Modern Distributed Tracing in .NET

By : Liudmila Molkova
Book Image

Modern Distributed Tracing in .NET

By: Liudmila Molkova

Overview of this book

As distributed systems become more complex and dynamic, their observability needs to grow to aid the development of holistic solutions for performance or usage analysis and debugging. Distributed tracing brings structure, correlation, causation, and consistency to your telemetry, thus allowing you to answer arbitrary questions about your system and creating a foundation for observability vendors to build visualizations and analytics. Modern Distributed Tracing in .NET is your comprehensive guide to observability that focuses on tracing and performance analysis using a combination of telemetry signals and diagnostic tools. You'll begin by learning how to instrument your apps automatically as well as manually in a vendor-neutral way. Next, you’ll explore how to produce useful traces and metrics for typical cloud patterns and get insights into your system and investigate functional, configurational, and performance issues. The book is filled with instrumentation examples that help you grasp how to enrich auto-generated telemetry or produce your own to get the level of detail your system needs, along with controlling your costs with sampling, aggregation, and verbosity. By the end of this book, you'll be ready to adopt and leverage tracing and other observability signals and tools and tailor them to your needs as your system evolves.
Table of Contents (23 chapters)
1
Part 1: Introducing Distributed Tracing
6
Part 2: Instrumenting .NET Applications
11
Part 3: Observability for Common Cloud Scenarios
16
Part 4: Implementing Distributed Tracing in Your Organization

Chapter 13 – Driving Change

  1. Using a single backend for all signals has certain advantages. It should be easier to navigate between signals: for example, get all logs correlated with the trace, query events, and traces together with additional context, and jump from metrics to trace with exemplars. So, using a single backend would reduce cognitive load and minimize duplication in backend-related configuration and tooling.

Using multiple backends can help reduce costs. For example, it’s usually possible to store logs in a cheaper log management system, assuming you already have everything up and running for logs and metrics. But these backends don’t always support traces well. Adding a new backend for traces and events only would make total sense.

Tools such as Grafana may be able to provide a common UX on top of different backends to mitigate some of the disadvantages.

  1. There are a few things that we need to do:
    • Lock down the context propagation...