Name node

• MOST IMPORATANT FOR CLUSTER!
• in-memory location of every Data block in cluster / rack
• Controlls DataNodes
• FS Ops
• Block Mapping

• Should have LOT OF RAM, LOT OF MEM, LOT OF POWER

• Map - Snapshot of entire file system (in Memory)
  • Tracks all the files
  • Tracks Replication value - can be configured by Replication factor
• Journal - keeps track back of Edit logs
  • Any client that changes/edits/adds information addes to Edit logs
• When proccess of Reboot or Checkpoint the Datanode marges the Map and Edit logs to fsImage

To keep presistnece of the data on hard drive the Edit log marges with Map