Jagane Sundar - BigDataNYC 2013 - theCUBE - #BigDataNYC
summary
WANdisco describes a concrete solution: Non-Stop software integrated into HDFS and HAWQ that provides up to seven NameNodes per cluster, uses Paxos-based distributed coordination (claims mathematically strong consensus instead of ZooKeeper shortcuts), and does asynchronous replication at the HDFS layer so distance is limited by bandwidth not latency. The product aims to deliver continuous availability across rack, site or whole data-center failures, and to eliminate NAS and load balancers while avoiding mapreduce startup delays.
AI-written summary · about
