Fixing Hadoop's Single Point of Failure With Veritas Cluster Server
Hadoop is an open source solution from Apache for managing and analyzing Big Data. Its scale-out architecture enables analyzing large volumes of data to find key business insights. Enterprises are turning to Hadoop for its agility and flexibility in data analysis. Hadoop enables enterprises make sense of varieties of data - structured to unstructured - and ask insight questions they were not able to do with traditional tools.
However Hadoop's Distributed File System (HDFS) has a weakness that makes it unattractive to enterprise datacenters. HDFS's meta data server, NameNode, is a single point of failure. When NameNode fails, applications lose access to data stored in many different DataNodes. In a long running analytical application a single failure can prevent enterprises from getting timely business insights.
Symantec's recommendation is to completely eliminate this flaw with a solution we are working on for enterprise ready Hadoop by making it highly available.
We recognize there are many work loads where enterprises prefer the scale-out model using many different commodity compute nodes with internal disks. But you do not have to settle for the single point of failure nor roll-out a solution that is not highly available. With our Veritas Cluster Server (VCS) you can make NameNode highly available and recover from failures in timely manner without operator intervention and without losing out on results.
Have an Hadoop installation with commodity nodes? Make it enterprise ready with VCS. See the detailed blog post our engineering team wrote on setting up NameNode in HA configuration using VCS.