I have tens of locations spread across the US, each with a varying number of DataMiner Nodes which are used by the local teams to manage their local operations; these DM nodes are part of the same DM cluster as there's a need to have an overview of all elements across the different locations.
What is the recommendation regarding storage architecture?
In this link I can't find a clear indication for this scenario Supported System Data Storage Architectures - DataMiner Dojo
Bruno - the storage architecture for distributed systems continues to be the same (Recommended setup: DataMiner, Cassandra, and Elasticsearch hosted on dedicated machines, with a minimum of three Elasticsearch nodes) as described in docs (Supported system data storage architectures | DataMiner Docs) following the correct compute requirements (DataMiner Compute Requirements - DataMiner Dojo).
When dealing with geographically separated nodes the attention is to be directed to network latency between the nodes, mainly. The general rule is to keep latency between 50 and 30 milliseconds, with special attention to the elastic nodes as specified in the documentation provided above. As, also seen, if fail-over is added to the equation, further constrains apply.
For complicated setups, it is always recommended to engage the architecture team to make sure the appropriate node distribution is applied.