Database Reference
In-Depth Information
The preceding diagram shows how different data storage level components are
organized. The
region server
contains the region
sales
, having two column families,
customer
and
orders
. Each column family has an in-memory storage, and single
column family records might have been written to multiple HFiles. In HBase, the
region server is treated as a datanode and supports multiple regions (represents
tables here). When data is written, by default, it goes to the
write ahead log
(
WAL
)
and the MemStore. HFile is created once the data present in the MemStore is lushed
to the disk and in case of lush failure, data is retrieved from WAL (we will discuss
all these components in detail in
Chapter 4
,
The HBase Architecture
).
Designing tables
In HBase, when modeling the schema for any table, a designer should also keep in
mind the following, among other things:
• The number of column families and which data goes to which column family
• The maximum number of columns in each column family
• The type of data to be stored in the column
• The number of historical values that need to be maintained for each column
• The structure of a rowkey
