I'm starting to test a very data-heavy data digestion process that writes to OpenTSDB over MapR-DB.
- When it starts off, OpenTSDB is empty, so the MapR-DB table is empty.
- As it writes, MapR-DB auto-splits its regions which I understand have to be between 4-6GB.
- So, as it runs more, I have more regions.
- My OpenTSDB process should distribute writes well as it uses salting, etc.
- Will my app write faster once there are more regions to write to? I have an 8 node cluster and each region seems to have one primary and 2 secondary nodes. This makes me think that I'll be writing to less servers in parallel while the region count is lower.
- Given regions in MapR-DB are 4-6GB, I guess you can't create them in advance like you can in HBase, right? The boundaries will change quickly as the regions grow in this case?