Hbase architecture and description

Hmaster: master node of the database

Hregionserver: A physical node is usually a single computer. An hregionserver contains multiple hregion. If a table contains 0.1 billion rows of data, it may be dispersed in different region of a regionserver.

Store: each column family is a store. when data is inserted, It is inserted in memstore. After a while, it is written into storefile by memstore. After a while, hbase starts a background thread, merge different storefiles to delete the rows whose flag is deleted.

Hfile: Each storefile is stored in only one hfile, and each hfile is distributed in the HFS file. Therefore, the storage is fully deployed.

Note: In hbase table design, each row is basically a column family with values and does not have values on two or more column families. The table design is sparse.