On GCP we provide two options to run enrichments: Beam Enrich, running on top of Google Dataflow and FS2 Enrich, running as a standalone JVM application.
Both applications that consume the raw data from the raw Pub/Sub topic (outputted by the collector). Validate the data (against schemas stored in Iglu Central or the user’s own schema registry(ies), enrich the data using one or more enrichments and then write the processed data out to the enriched Pub/Sub topic, from where it can be e.g. loaded into BigQuery.
FS2 and Beam Enrich are accessible as a docker images (from dockerhub). You can also build the container yourself from source with
sbt "project beam" docker:publishLocal (or
sbt "project fs2" docker:publishLocal) and build the archive from source using
Both options provide same functionality but with different performance/management trade-offs. Beam has to be deployed as a Dataflow job and provides good performance and auto scaling for very big volumes of data. In some cases though we don’t need this high throughput and Dataflow is an expensive and opaque service. In these cases you can use Enrich FS2 which is much cheaper for low volume pipeline and easier to manage in absence of scalability (although it also can be scaled using Kubernetes or similar orchestration tool)