Tyrannical 9s solved with OpenTracing, your Apps and Couchbase

As was outlined in the Tail at Scale The Tail at Scale[1], the math is against us on distributed systems.  Seemingly momentary spikes do not just cause a small blip in your service level.  On a distributed system, they can multiply and stack up, leading to a declining user experience as services are broken into smaller and smaller pieces.

OpenTracing is one solution for this.   One of the top-level CNCF projects, OpenTracing is an set of vendor neutral APIs for distributed tracing.  OpenTracing defines a set of interfaces and lets developers add tracing to key points in their system architecture to be able to understand how a request flows through the system.  

Using OpenTracing based systems and instrumentation from other platforms like databases, we developers can build systems that have observability into where the pauses are, what goes into a component, and how that impacts service levels. 

In this talk, Matt Ingenthron from Couchbase will expand on the challenge, introduce OpenTracing and the concepts behind it, and then talk about how Couchbase integrated OpenTracing into the system.  Couchbase provides both out-of-the-box tracing and can be plugged into OpenTracing compatible Tracers such as Jaegar, LightStep, DataDog, OpenZipkin or other systems.

With this, one can build caching or data services into services and finally see where the time goes across the many requests flowing from the app, to the database API, to the network, to the backend and back.  The same technique can be used across your own microservices.

1. The Tail at Scale, Communications of the ACM, Volume 56 Issue 2, February 2013 

Pages 74-80, Jeffrey Dean, Luiz André Barroso