The data has to be split into smaller, manageable chunks before loading. But it is suggested to first discuss with the configuration manager, CI class owners, service owners, application owners etc. to make sure that the target classes are correct. If all the 4M CIs are going into only 2 CI classes, there is a high possibility of many CIs getting populated in wrong CI classes (unless this is already fully validated and confirmed by relevant stakeholders). Exclude already inactive CIs from the data load as much as possible. Also if you do not have a plan to automate the data updates for most of these records in the future (via discovery or integrations), it is not recommended to load so much data as the data will become stale soon without any automation to update & maintain it (4M is a very huge number which would be impossible to maintain manually).
