As Spark avoids performing heavy disk I/O, it has a significant improvement in performance over Hadoop.

• e.g. PageRank time

• e.g. other iterative algorithms
  • Spark tries to keep data that will be used again in memory, as opposed to reading/writing from disk