Post by Admin on Mar 14, 2014 18:51:11 GMT
1. Replicate dimension tables from transactional databases with low latency and without sharding. (Fact data can stay in Hire since it is append only)
2. Only more changed tows.
“Full scrpe” is too slow does not scale as data growing.
Hire by itself is not good at
3. Integrate into Hire’s map reduce query execution plans for parallel distributed processing.
4. Multi versioning for snapshot consistency.