Deployment Options
Deployment Topography
The suggested deployment configuration, if both ODS Inquiry and ODS Ingestion are required, is 3+ ods-ingestion instances, and 1+ ods-inquiry instances.
ODS Ingestion is stateful, and forms an Akka Cluster, which requires a minimum of three nodes.
ODS Inquiry is stateless, and should be scaled as required. One instance is typically enough.
IPF Archiver requires ODS Ingestion to be running, and performing archive candidate selection.
IPF Metrics Processor requires ODS Ingestion to be running because in some scenarios it will need to query the ODS database.
| If ODS Ingestion is running to support IPF Archiver or IPF Metrics Processor only, ODS Inquiry won’t be needed, and the number of ODS Ingestion nodes can be decreased to one. |
Both will use the same MongoDB database. Ingestion writes and updates data, and Inquiry queries this data.
Indexing
Indexes come with both a write performance cost, and storage cost, with the upside being faster and cheaper queries.
Indexes that support ingestion are mandatory, and all ODS Ingestion indexes should be applied, but many inquiry indexes are not necessary in all cases.
Depending on requirements for ODS Inquiry, some or all ODS Inquiry indexes should be applied. It is recommended that all inquiry indexes for the summary collection be applied, as these are required to support GUI searches.
Indexes required for inquiry on other collections may not be required, and depend entirely on your specific requirements.
ODS will create all indexes on startup, but this can be disabled. See the indexing section for more information.