Voxxed Days Berlin is right around the corner! We sat down with speaker Adrian Cole from Pivotal to
discuss his upcoming talk, ‘How to Properly Blame Things for Causing Latency: An Introduction to Distributed Tracing and Zipkin.’
Adrian is an active member of cloud interoperability circles. He is the founder of a few popular open source projects, notably Apache jclouds and Netflix Feign. Recently, he’s focused on distributed tracing, particularly in the OpenZipkin project. Adrian works at Pivotal in the Spring Cloud OSS team.
Voxxed Days Berlin: You state in your talk abstract that latency analysis is a “blame game.” How so?
Cole: I’m half-joking, but the other half is I suppose what you are asking. When we troubleshoot things, we fill gaps of signal with experience and hunches. For example, if I look at lines in a log file and see a gap of 30 seconds, I’m left to guess what might have happened. Who knows! Maybe I blame it on Garbage Collection? Or high CPU? Or that new service just deployed? Even with aggregate metrics, it is hard to navigate signals that often end up in logs, and some of that ends up as hypothesising (or blaming) systems based on our experiences with them.
Why is distributed tracing important for microservice architectures?
Microservices are decomposed across multiple processes. Latency is a common concern, and distributed tracing is special cased for service granularity latency troubleshooting.
In Zipkin, for example, you can query for any operations (spans), which took longer than a second. What you get back is a latency graph, a tree of your service calls. This helps you get to the bottom of things.
Speaking of which, can you please tell us a bit more about the open source project Zipkin and who is contributing to it?
Zipkin owes credit to its original founders, Twitter, and the Google Dapper paper on which it was based. In the last year Zipkin left the nest (pun intended) and is now an organisation called OpenZipkin.
To list contributors means me leaving many out by mistake! OpenZipkin has had a lot of recent work from Uber, MdSol, SoundCloud (here in Berlin), Yelp, Prezi, Firebase and my team: Spring Cloud at Pivotal.
We are also really encouraged to see work from OpenTracing (OT), something I participate in. OT is an effort to standardize how tracing systems can interoperate. For example, how you can choose between tracing backends. It is similar to SLF4J, except polyglot and yeah for distributed tracing.
Join 300 developers at Voxxed Days Berlin on 28-29 January. Over two days of conference, you’ll learn from Rock Star speakers like Neal Ford, Ted Neward, Sven Peters, Juergen Hoeller, Simon Maple, and more. Even better, Voxxed.com readers can claim a 30% discount on their tickets. ‘Use “VXB16_VXD_g83nK4” and save!