Zipkin is a distributed tracing system that helps us gather timing data for all the disparate services at Twitter.
You can find the code in the "zipkin-aggregate" submodule. In our setup, it runs every night at 3 AM. The job reads data directly from Cassandra, analyses it with a Scalding/Hadoop job and then writes the results back into Cassandra. This aggregate data can then be displayed in the "Aggregate" UI.
We draw a graph of all services in the platform, built from the aggregated data. This graph is browseable and shows metadata about each service/dependency.