This presentation was given by Bernd Ocklin, Senior Director MySQL Cluster Development at Oracle.  It was a good overview of what NDB Cluster is, where it came from and some recent changes.  My notes are:

• Trade-Offs: Cellular Network
  • used as backend for huge telecoms
  • http://bit.ly/oR15tF – for an example in use
  • massive volumes
• Trade-Offs: Online Games – Massive Parallel
• 99.999% uptime
• Distributed real-time in-memory network database
• Runs standalone or as a storage engine for MySQL
• Auto-sharding based on key
  • transparent between node groups
  • each fragment active in one Data Node with synchronous replication to 2nd data node
• DBA chooses what part of Primary key to shared on
• parallelization works even on single queries
• batches automatically
• event-driven and asynchronous
• Active-Active with Geo-Replication in NDB Cluster
  • Update anywhere
  • conflict detection
  • auto-conflict-resolution
• Async-Replication is for read scale out
• MySQL Group Replication – innoDB Cluster
  • scale out writes as well as reads (limited write scale out)
• Moving towards mixing NDB cluster as a node in InnoDB cluster (not there yet)
• Synchronous locally and use replication for geo-replication (asynchronous)
• Updates scale pretty much linearly
• version 7.6 new features
  • makes joins much faster
  • added ndb import to read in csv files
  • MEM integration
  • MySQL 5.7
  • dynamic resource allocation
  • Partial Local Checkpoint – for performance improvements and scalability (now support much larger databases)
  • Multi-treaded backup
• in MySQL 8.0 they are moving NDB to base MySQL to avoid forking and placing it as a first class citizen
• MySQL 8.0 new feature
  • CTEs
  • Window Functions
  • Data Dictionary
  • utf8mb4
  • Security Roles
  • Improved Information Schema Performance