3 Sweet Spots for MapReduce

by Alex Popescu
Keywords: mapreduce,NoSQL, NoSQL databases, NoSQL community, the definitive guide to NoSQL

Twitter Reddit
22:52

The presentation given by Andrew Pavlo “MapReduce and Parallel DBMSs”, embedded below for reference, identifies the following 3 sweet spots for MapReduce:

  • Extract-Transform-Load
    • “Read Once” data sets
    • Allows for quick-and-dirty data analysis
  • Semi-Structured Data
    • Can easily store semi-structured data which would otherwise be awkward to be stored in RDBMS
  • Limited Budget Operations
    • the alternative, parallel DBMSs are expensive

When speaking about the possible MapReduce and RDBMS integration, something that for example Oracle has already been considering, Andrew and his colleagues mention the following advantages:

What can MapReduce learn from Databases?

  • Fast query times.
  • Schemas.
  • Supporting tools.

What can Databases learn from MapReduce?

  • Ease of use, “out of box” experience.
  • Attractive fault tolerance properties.
  • Fast load times.