Write Path
HBase is designed to optimize writes. This is achieved by splitting the load across all the Region Servers. The design is optimized to handle Region Server failure and data loss. HBase data is orchestrated very similarly to store map, such as sorted key space partitions located in different shards/regions.
HBase tables are managed by a series of servers, as follows:
- One active master server
- Multiple backup master servers (you can use one backup too)
- Many Region Servers
When a Region Server receives a write request, it relays the call to the specific region. Each region stores sets of rows. Row data can be separated into multiple column families (we will discuss this in the chapter about schema design). The data of a particular column family is stored in HStore, which has Memstore and a set of HFiles. Memostore is kept in the Region Server main memory, during which the HFiles are written to HDFS.
Note
Note that you can plan to flush the changes in the form of batches by switching off...
