I wish I had the same problem, I am working on presenting 1M data as clusters and trying to solve problem with superclusters but 1M data creates serious problem. Withbounding box approach we can present 20000 data on map within 45 seconds but still it is not the main target