Skip to main content

What is end-to-end distributed tracing?

Modern apps are increasingly composable—meaning they are made up of multiple interconnected services or microservices. A single user request follows a complex path across these distributed components. End-to-end tracing follows each transaction request from the source and into the depths of your multi-cloud and datacenter infrastructure. When a transaction is stitched end-to-end, you can see exactly where the delays and errors in your application lie and fix them faster.

Most modern application performance monitoring (APM) solutions provide some form of distributed tracing, but not all of them capture every transaction or stitch them end-to-end.

Why is complete data important for distributed tracing?

Complete end-to-end distributed tracing is essential for global application analysis and comprehensive code profiling.

Here are 5 reasons why:

1. You need to capture and store every single transaction so that you can reconstruct them for later analysis and troubleshooting. This helps you track down and address even transient or intermittent problems the FIRST time they occur.

2. You need global application data to see the true impact of transaction latency at the component level. The same component could be used by multiple transactions and called hundreds of times a day. Unless you have this global view you won’t know that a seemingly innocuous 53 ms method is causing 700 hours of latency over the course of a day!

3. You need to trace EVERY transaction and method in its call stack, not just on the tiers that exceed thresholds. Unless you do this, that troublesome 53 ms method will fly under the radar.

4. You need to troubleshoot the long tail of performance issues, not just the biggest cause. A single symptom can have 11 different contributing factors and all of them need to be investigated. Any one of those factors could potentially wreak havoc elsewhere in the application environment.

5. You need to understand cross-application and cross-component dependencies to properly allocate and optimize resource usage. A critical app could be starved of connection pool resources by a well-performing, little-used app that doesn’t trigger any thresholds—how would you know if you rely on snapshots or sampling-based approaches?

Complete data includes both monitoring and observability-based data sources. In addition, big data technology is necessary to ensure you always have all the evidence you need without overly taxing your applications in production or incurring excessive storage expenses.

With complete end-to-end distributed tracing, you can be confident that your applications are properly optimized for peak usage periods and other real-world scenarios.

What are some key use cases for distributed tracing?

With end-to-end distributed tracing, you can capture every transaction for diagnosis, fully understand application dependencies and get insights into users’ behavior.

Use distributed tracing data to:

  • Map your application flow across your infrastructure for each and every unique transaction.
  • Find and fix vulnerabilities quickly every time you release new features or change configuration.
  • See which operations are popular and which ones are not to properly prioritize new features, issue resolution and infrastructure changes based on usage patterns.
  • Statistically analyze performance data of all transactions. Study outliers and compare against baselines.
  • Understand global patterns and dependencies across your dynamic multi-app, multi-cloud environment.
  • Optimize cloud costs by thoroughly profiling your applications in dev/test.

The support team spent 90% of its time trying to figure out what was going wrong with certain applications.

Mark Osborne
Web Systems Engineer
National Instruments

Vous êtes prêt ?

Exploitez pleinement le potentiel de vos investissements dans le digital avec Riverbed