WANdisco today announced its Non-Stop NameNode, a 100 percent uptime solution for Hadoop. Non-Stop NameNode eliminates the single point of failure that risks data availability — the NameNode — which tracks the data's location within a Hadoop cluster and coordinates client access to it.
"If the NameNode is unavailable, the Hadoop cluster goes down," said David Richards. "With other solutions, a single NameNode server actively supports client requests and complex procedures are required if a failure occurs. The Non-Stop NameNode eliminates those issues and also allows for planned maintenance without downtime. WANdisco provides 100% uptime with unmatched scalability and performance."
Built on the company's Non-Stop technology for deployment on Apache Hadoop 2, Non-Stop NameNode enables multiple active NameNode servers to act as peers, maintaining continuous synchronization and simultaneously supporting client access to the Hadoop cluster. If one server fails, or is taken down for scheduled maintenance, others take over automatically without any downtime. When the server comes back up, it immediately resynchronizes itself with the others in the cluster and resumes normal operation. This eliminates the risk of downtime and data loss due to human error during failover and recovery.
"Hadoop was not originally developed to support real-time, mission critical applications, and thus its inherent single point of failure was not a major issue of concern," said Jeff Kelly, Big Data Analyst at Wikibon. "But as Hadoop gains mainstream adoption, traditional enterprises rightly are looking to Hadoop to support both batch analytics and mission critical apps. With WANdisco's unique Non-Stop NameNode approach, enterprises can feel confident that mission critical applications running on Hadoop, and specifically HBase, are not at risk of data loss due to a NameNode failure because, in fact, there is no single NameNode."
Following are some key features of the Non-Stop NameNode:
- Every server is a continuous hot backup of every other, providing automated failover and recovery after server outages without administrator intervention.
- Non-Stop NameNode automated failover and recovery features can be used to eliminate downtime for scheduled maintenance.
- Risk of data corruption due to "split-brain" is eliminated because all NameNode servers actively support clients and are always in sync.
- Non-Stop NameNode scales as more servers are added to the cluster to support increased workload without sacrificing performance.
- Non-Stop NameNode delivers the same capabilities over a WAN regardless of distance.
- Provides full support for HBase and works with Hadoop 2, WANdisco Distro (WDD) and CDH 4.1.x.