HBase Vs. HDFS and Architecture of HBase

HBaseVs. HDFS and Architecture of HBase

HDFS

HDFS is a simple distributed file system that is well suitable for storing larger files. It is a general purpose file system and so is not for the individual record lookups to be done in the files.

HBase (  HBase Training )

HBase provides fast record lookups as it is built on top of the HDFS, especially for the larger tables. For high–speed lookups, the data is put internally with ‘StoreFiles’, in indexed form, which would exist in HDFS. The files would be sequentially accessed in HDFS .

HBase Architecture

HBase is written over the file system of HDFS. It can be accessed through native Avro, Thrift, Java API, REST etc. Zoo Keeper here works as the coordination server to get the entire system to be synchronized and the bootstrap. 

There are three pieces in HBase

1.      Master – Master is only one here.It manages the cluster.

2.    RegionServer – RegionServers can be multiplied. Its function is to manage the data. It again has three components.

a.      HFile

b.      Memstore

c.       Write-Ahead Log

3.      API – API is meant to automatically pick the current value present in the cell, by default.

In HBase, the basic things to be known are:

1.    Column – It is the most basic unit, containing a Key Value. Number of columns is fixed, unless flexible schema is present. Here each row can have different columns. These columns can have multiple versions, in which the distinct values are contained in the cells. The columns often are referred as family qualifier. And the qualifier is considered as any of the arbitrary array of bytes.

2.      Row - When it comes to a row, it would be a collection of column families. These column families are like Personal Info, Banking Info and Career Info etc. Vacant columns can be simply ignored. The rows are sorted lexicographically using their row key. Lexicographical sorting means, each of the specified keys would be compared on a binary level bit by bit, byte by byte and from left to right. Here, rows are sequentially sorted like Row1, Row2 etc.

3.      Table - Unlike traditional table model, a cell has a list or collection of name values for the specified id. Multiple values for example of ‘Personal Info’ can be contained in the one column family.

4.      Filter - HBase does not entertain concept of the query, instead it uses the filter to filter the data and access. Two filters are used. These filters are server side filter and the other one is the client side filter.

Conceptual model:

According to the conceptual model, we can see how the RDMBS model is converted to the HBase model and how the Rows and Columns are treated together.

By seeing the conceptual model, we can understand that HBase does make the data to be easily handled by shrinking or shorting multiple Tables into one table, since the columns are treated as the column families. Each column family can contain the columns of a Table. Complex queries with Joint can be minimized and made easy.

For Apache Hbase online Training and Hadoop Hbase Tutorial, you can visit Intellipaat.uk.

To access the data, it needs the expression of

(Table, RowKey, Family, Column, Timestamp) à Value