There are multiple options for tracing (Zipkin, Dapper, and others) but a few of them seem to have become mainstream. One of them is the OpenTracing initiative (http://opentracing.io/). All share more or less the same design based on spans.
The global idea is to let each transaction’s actors mark their presence with a span. A span contains an identifier, some metadata about the invocation, and the execution time. The identifier is generally composed of multiple values representing the overall trace identifier (the request marker), the span identifier, and, often, the parent identifier.
When correctly installed, the tracing happens on the client and server sides, so you have a full vision of the system handling, and it is associated ...
