Database Reference
In-Depth Information
Following are a few important characteristics of Hadoop framework:
• Runs on commodity servers, with elastic scalability; the storage servers and
disks are not assumed to be highly available or reliable
• Uses replication strategy to address high availability and reliability require-
ments
•
Metadata driven design
: Master server called NameNode holds the
metadata and the slave nodes called DataNode store data and can handle
data processing as well
• Focus here is mostly sequential access with single writes and limited or no
file locking features
• Comes with an in-built foolproof fault-tolerant and high-availability architec-
ture
Hadoop Distributed File System (HDFS)
The figure below depicts HDFS architecture with the associated daemons:
HDFS architecture follows master-slave architecture paradigm. The master node is
referred to as NameNode and has the following functions:
• File system namespace management
• File name to list blocks and location mapping details maintenance
• Block allocation/replication management
• Checkpoints namespace and journals namespace changes for reliability
• Access control to namespace
Search WWH ::
Custom Search