concepts

• Solve the problem of iterative algorithm
• MR pb: repetead access to HDFS, no data caching in between iteration
• MPI: no support for fault tolerance
• What does spark
  • Iterative algorithm
  • Interactive data mining
  • Use RDD
  • Keep features of MR: fault tolerance, data locality, scalability
• Data Model
• RDD
• Logistic regression ?