Just found this interesting presentation,
Map-Reduce-Merge: Simpli?ed Relational Data Processing on Large Clusters
by Hung-chih Yang, Ali Dasdan Ruey-Lung Hsiao, D. Stott Parker; as presented by Nate Rober (PDF)
Excerpts:
Extending MapReduce
1. Change to reduce phase
2. Merge phase
3. Additional user-de?nable operations
a. partition selector
b. processor
c. merger
d. con?gurable iteratorsImplementing Relational Algebra Operations
1. Projection
2. Aggregation
3. Selection
4. Set Operations: Union, Intersection, Difference
5. Cartesian Product
6. Rename
7. Join
[for more detail see full slides]
Conclusion
MapReduce & GFS represent a paradigm shift in data processing: use a simpli?ed interface instead of overly general DBMS.
Map-Reduce-Merge adds the ability to execute arbitrary relational algebra queries.
Next steps: develop SQL-like interface and a query optimizer.
Research paper: Map-reduce-merge: simplified relational data processing on large clusters (PDF for ACM people)
Linked from HRDF page in the Hadoop wiki, where there appears to be a proposal brewing to build an RDF store on top of the Hadoop/Hbase infrastructure.
Nearby: LargeTripleStores in ESW wiki
Not entirely unrelated: Google Social Graph API (which parsers FOAF/RDF from ‘The Web’ but discards all but the social graph parts currently)