Is having long-running analytical queries a bad thing on a highly transactional workload?
Is running them on a Replicated database a better idea?
As with any database, resources are limited in SingleStore by the size and configuration of the available hardware. Therefore, analytic queries in SingleStore will be default leverage as many compute resources as possible to optimize performance. However, this can be problematic if concurrent transactional queries with tight SLAs are vying for those same resources. The traditional approach to the scenario is via resource governance, which is possible in Singlestore; however, it may provide another option if a DR cluster is deployed.
A DR cluster deployed using intra-cluster replication effectively persists an independent copy of the entire dataset, which is available for read-only querying. Executing long-running analytic queries against the DR cluster therefore allows them to run without creating resource contention on the primary cluster that may impact SLAs for other queries.